v2.0.5: FAST_MODEL routing + tool-use summaries + Qwen dispatch + Arena

Source-level recon of QwenLM/qwen-code (Apache-2.0) informed 4 lifts: 1. FAST_MODEL config: optional env var routes cheap LLM calls (titles, summaries, labeling) to a smaller model on llama-swap. auto_name.ts uses ctx.config.FAST_MODEL ?? session.model. Set FAST_MODEL=nemotron- nano-4b to avoid loading the 35B model for 20-token title generation. 2. Tool-use summaries (services/inference/tool-summaries.ts): utility that generates "git-commit-subject-style" labels for tool batches via a fast-model LLM call. System prompt + truncation logic ported from Qwen Code's toolUseSummary.ts. Exported via @boocode/server/inference for BooCoder's dispatcher to call after task completion. 3. Qwen as dispatchable agent: added to agent-probe.ts KNOWN_AGENTS. PTY dispatch builds: qwen -p "<task>" --output-format stream-json (NDJSON structured events over stdout). Env: OPENAI_BASE_URL + OPENAI_API_KEY points Qwen Code at llama-swap. execution_path CHECK constraint extended with 'qwen'. 4. Arena routes (routes/arena.ts): POST /api/arena dispatches the same task to N contestants (2-5, each with different agent/model), each getting its own task row linked by arena_id UUID. GET /api/arena/:id shows all contestants. POST /api/arena/:id/select/:task_id marks winner. Schema: arena_id column added to tasks. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
v2.0.4-hardening: fuzz suite + integration tests + production readiness
2026-05-25 14:05:59 +00:00 · 2026-05-25 04:31:22 +00:00 · 2026-05-25 04:25:18 +00:00 · 2026-05-25 04:17:28 +00:00 · 2026-05-25 04:10:46 +00:00 · 2026-05-25 03:55:18 +00:00
160 changed files with 13471 additions and 625 deletions
--- a/.gitignore
+++ b/.gitignore
@@ -10,3 +10,4 @@ secrets/
 data/*
 !data/AGENTS.md
 !data/skills/
+!data/mcp.json
--- a/BOOCHAT.md
+++ b/BOOCHAT.md
@@ -28,10 +28,24 @@
 - Prefer codecontext (`search_symbols`, `get_symbol_info`, `get_dependencies`) over `grep` for symbol-level questions. Fall back to `grep` / `view_file` when codecontext returns degraded or empty results — that signals an unsupported language or parse failure.
 - Verify before reporting work complete: run the relevant test/build/smoke command and confirm output matches the claim. Evidence first, assertion second.

+## Output format
+
+- Stay in Markdown by default for every reply, short or long.
+- Switch to a self-contained `<!DOCTYPE html>...</html>` artifact only when the user explicitly asks (e.g. "render this as HTML", "make me a dashboard", "build an interactive diagram"). Detection is opportunistic — the BooChat backend tags the assistant message as an HTML artifact, opens it in a sandboxed pane, and offers Download. Do not emit HTML unprompted; long Markdown is the right answer for most explanatory output.
+- When asked to produce HTML, avoid generic AI aesthetics: no excessive centered layouts, no purple gradients, no uniform rounded corners, no Inter font. Prefer interactive controls (sliders / knobs / SVG / side-by-side diffs) over passive prose-in-HTML. Pattern reference: claude.com/blog/using-claude-code-the-unreasonable-effectiveness-of-html (Thariq Shihipar, May 2026).
+- The HTML artifact is rendered in a sandboxed iframe with `connect-src 'none'` — `fetch()`, WebSockets, and tracking pixels do not work. All logic must be client-side.
+
 ## Convention: rules vs recipes

 Always-true rules (process discipline, refusals, behavior contracts) live here in `BOOCHAT.md` — and in `BOOCODER.md` / `CLAUDE.md` per their scopes — where they are 100% present in every turn. On-demand recipes (specific procedures, scaffolds, checklists) live in `/data/skills/` and invoke roughly 6% of the time in clean multi-turn flow (Codeminer42 measurement, 2026). Don't file workflow rules as skills — they silently misfire. See Anthropic agent-skills best-practices (platform.claude.com/docs/en/agents-and-tools/agent-skills/best-practices) for the canonical conventions.

+## Verification discipline
+
+- When assessing implementation status, verify against the running container (`curl /api/health`) and latest git commit (`git log --oneline -3`), not just source file contents. Source files can be mid-edit. The deployed state is the truth.
+- Never count `dist/` directory sizes as source lines. Only count `src/**/*.ts` files. Compiled output is inflated by inlined types and transpilation artifacts.
+- Before claiming a feature works, run the actual command and show the output. "Should work" is not verification. Acceptable evidence: test output (`pnpm test`), build output (`pnpm build`), curl response, docker logs, `\d tablename` output. If you can't run it, say so explicitly — don't assert success without evidence.
+- When reporting counts (tools, tests, files, routes, lines), derive the number from a command (`grep -c`, `wc -l`, test runner output) — not from memory or approximation.
+
 ## Known limitations

 - Codecontext re-analyzes the project graph on each call against a different target_dir. First call to a new project may take 1-3 seconds; subsequent calls to the same project return in ~10ms.
--- a/BOOCODER.md
+++ b/BOOCODER.md
@@ -1,27 +1,39 @@
-# BooCoder
+# BooCoder — Container Guidance

-> (Stub. v2.0 implementation pending. This file documents the intended contract.)
+You are BooCoder, a write-capable coding agent. You can read AND modify files within the project scope.

-## Capabilities
+## You can

- Everything in `BOOCHAT.md`
- Write tools (pending): `write_file`, `edit_file`, `delete_file` (all gated through pending-changes sandbox)
- Shell (pending): `run_command` (Docker-isolated per-session)
+- Read files (view_file, list_dir, grep, find_files)
+- Edit files (edit_file, create_file, delete_file) — all changes queue in pending_changes
+- Apply pending changes to disk (apply_pending)
+- Revert applied changes (rewind)
+- Dispatch tasks to external agents (dispatch_external_agent)
+- Use MCP tools from configured servers

-## Constraints
+## You cannot

- All writes land in a pending-changes virtual layer; nothing touches the real filesystem until `/apply`
- `run_command` executes inside the session sandbox, not the host
- No git commits, pushes, or pulls — Sam owns those
- Stop and ask before destructive operations (delete, overwrite, recreate)
+- Write outside the project root (path-guard enforced)
+- Write to secret files (.env, *.pem, id_rsa*, credentials.json)
+- Apply changes without explicit user approval (unless auto-apply is enabled per task)
+- Push to git remotes
+- Access the internet except via configured MCP servers
+
+## Pending changes discipline
+
+Every file modification queues in `pending_changes` before touching disk. The user sees a diff preview and approves/rejects each change. Never bypass this queue — it is the safety boundary between inference and the filesystem.

 ## Behavior

- Show a diff preview before any write
- Group related edits into a single `/apply` batch
- If a tool fails, surface the error verbatim — don't paper over it
+- Show diffs clearly. Explain what you're changing and why.
+- For multi-file changes, organize as a logical unit (one task = one coherent change set).
+- If uncertain about scope, use smaller edits and verify between steps.
+- Cite file paths + line numbers for context.
 - Verify before reporting work complete: run the relevant test/build/smoke and confirm output matches the claim. Evidence first, assertion second.

-## Convention: rules vs recipes
+## Verification discipline

-Always-true rules live here, in `BOOCHAT.md`, and in `CLAUDE.md` (100% present each turn). On-demand recipes live in `/data/skills/` (roughly 6% invoke rate in multi-turn per Codeminer42, 2026). Don't file workflow rules as skills — they misfire. See Anthropic agent-skills best-practices (platform.claude.com/docs/en/agents-and-tools/agent-skills/best-practices).
+- When assessing implementation status, verify against the running container (`curl /api/health`) and latest git commit (`git log --oneline -3`), not just source file contents. Source files can be mid-edit. The deployed state is the truth.
+- Never count `dist/` directory sizes as source lines. Only count `src/**/*.ts` files. Compiled output is inflated by inlined types and transpilation artifacts.
+- Before claiming a feature works, run the actual command and show the output. "Should work" is not verification. Acceptable evidence: test output (`pnpm test`), build output (`pnpm build`), curl response, docker logs, `\d tablename` output. If you can't run it, say so explicitly — don't assert success without evidence.
+- When reporting counts (tools, tests, files, routes, lines), derive the number from a command (`grep -c`, `wc -l`, test runner output) — not from memory or approximation.
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -2,6 +2,42 @@

 All notable changes per release tag. Most recent on top, ordered by tag creation date (which matches the git history). Tag names follow `vMAJOR.MINOR.PATCH-slug` — the slug describes what shipped, so the tag name alone is enough to recall the batch.

+## v2.0.4-hardening — 2026-05-25
+
+Path-guard fuzz suite: 25+ traversal-attack tests covering ../ sequences (all depths), encoded traversal (%2e%2e), null byte injection, absolute path escape, prefix-without-separator, backslash traversal, and the full secret-file deny list (.env, *.pem, id_rsa*, *.key, credentials.json, *.kdbx, .netrc). Plus 5 valid-path positive tests confirming normal writes aren't blocked and 5 edge-case tests (empty, whitespace-only, very long path, triple-dot, multiple slashes). Null-byte and whitespace-only guards added to `resolveWritePath` (previously only checked empty string). DB-integration test skeleton for pending_changes full-cycle (queue create/edit/delete, apply, rewind) gated on DATABASE_URL via `describe.runIf`. Production readiness verified: all services healthy, all builds clean, 57 tests passing (23 existing + 34 new).
+
+## v1.16.0-codesight-merge — 2026-05-24
+
+Ports codesight's highest-value analysis capabilities into the codecontext sidecar as 4 new MCP tools. Tier 1 (graph queries on existing edges, no re-parsing): `get_blast_radius` (BFS reverse-edge traversal — "what breaks if I change this file?", with depth tracking) and `get_hot_files` (most-imported files ranked by incoming edge count — change-risk indicators). Tier 2 (tree-sitter AST re-parsing on demand): `get_routes` (Fastify/Express HTTP route extraction with method, path, file, line, inferred tags for db/auth/cache) and `get_middleware` (middleware registration detection via import-name heuristics and app.register/addHook/setErrorHandler patterns, classifying as auth/cors/rate-limit/security/error-handler/logging/validation). All 4 tools use `defer s.graphMu.RUnlock()` for consistent mutex discipline (reviewer caught that the initial implementation released the lock early on the Tier 2 tools). Route object-property extraction delegates to `extractStringValue` for template-literal handling (reviewer catch). codecontext sidecar rebuilt from `/opt/forks/codecontext` commit `b19e646`, tagged `v1.16.0-codesight-merge`. BooCode wrapper tools follow the existing codecontext pattern — 4 new files in `apps/server/src/services/tools/codecontext/`, registered in ALL_TOOLS. 29 new Go tests + 363/363 BooCode server tests passing. No schema changes, no frontend changes.
+
+## v1.15.0-mcp-multi — 2026-05-24
+
+Multi-server MCP client with stdio + Streamable HTTP transports, JSON config file, and per-agent tool glob patterns. Generalizes the v1.14.1 single-server Context7 PoC into a registry of named MCP servers with per-server graceful degradation. JSON config at `/data/mcp.json` (bind-mounted alongside `AGENTS.md`) matches opencode's `mcpServers` schema shape so server entries are copy-pasteable. Config file missing = no MCP (opt-in by file presence). Stdio transport spawns a persistent subprocess via the SDK's `StdioClientTransport` with NDJSON framing; Streamable HTTP reuses the v1.14.1 pattern via `StreamableHTTPClientTransport`. Tool prefix generalized from `context7_<name>` to `<serverName>_<toolName>` with a reverse `toolToServer` map for dispatch routing. Per-agent AGENTS.md `tools:` field now supports glob patterns (`context7_*`, `!web_*`) via `matchToolGlob` (last-match-wins, `!` prefix denies); replaces the exact-match `.includes()` in `stream-phase.ts`. Glob patterns bypass `ALL_TOOL_NAMES` validation in the parser since MCP tool names aren't known at parse time. `refreshToolNames()` in `agents.ts` rebuilds the `DEFAULT_TOOLS` snapshot after `appendMcpTools` so agents without explicit `tools:` lists see MCP tools — reviewer caught that the module-load-time snapshot would permanently exclude late-registered tools. Read-only invariant preserved: all MCP tools with `readOnlyHint: false` rejected at discovery. Result size capped at 5MB. Shutdown hook closes all transports. v1.14.1 env vars (`MCP_CONTEXT7_URL`, `MCP_CONTEXT7_API_KEY`) removed — superseded by the config file. Default `data/mcp.json` ships with Context7 disabled; flip `"enabled": true` to activate. 363/363 server tests passing (27 new: multi-server wrapping, glob matching, routing, degradation). No schema changes, no frontend changes.
+
+## v1.14.1-mcp-poc — 2026-05-23
+
+Single-server MCP client PoC against Context7. New `apps/server/src/services/mcp-client.ts` (~200 lines) wraps `@modelcontextprotocol/sdk` v1.29.0 with Streamable HTTP transport. On startup (when `MCP_CONTEXT7_URL` is set), connects to Context7, discovers tools via `tools/list`, wraps each as a `ToolDef` prefixed `context7_<name>`, and appends to `ALL_TOOLS` (alpha-sorted for prompt-cache stability). `appendMcpTools()` in `tools.ts` handles the late-registration; `ALL_TOOLS` changed from `ReadonlyArray` to mutable to support it. Read-only invariant guard rejects any MCP tool with `readOnlyHint: false` (MCP SDK v1.29.0 uses `readOnlyHint`, not `readOnly`). Tool dispatch is transparent — `executeToolCall` routes MCP tool calls through the `ToolDef.execute` wrapper, which strips the `context7_` prefix before calling the MCP server. Graceful degradation: MCP server down at startup → zero tools, warn log; MCP server down mid-session → error-shaped result, model self-corrects. Result size capped at 5MB with truncation (matches native `view_file`'s `MAX_FILE_BYTES`). Adversarial review caught that the Zod `.default('https://...')` on the URL config made MCP effectively always-on instead of opt-in — fixed by removing the default. 348/348 server tests passing (16 new mcp-client tests covering tool wrapping, read-only guard, name prefixing, content extraction). No schema changes, no frontend changes. Proves the MCP tool-discovery → tool-call → result-render loop end-to-end before the full v1.15 port.
+
+## v1.14.0-outer-loop — 2026-05-23
+
+Converts the inference engine's ad-hoc `executeToolPhase → runAssistantTurn` recursion into an explicit `while` loop with a configurable step cap. A step is one stream-and-tool-execute iteration; the loop terminates on non-tool finish, step-cap hit, doom-loop, budget exhaustion, abort, or synthesis success. `MAX_STEPS = 200` is the hard ceiling (4x the old effective limit from budget); per-agent `steps:` field in AGENTS.md frontmatter sets tighter caps (Refactorer: 5, Architect: 20, others: unset = bounded only by MAX_STEPS). `executeToolPhase` no longer recurses — returns a `ToolPhaseResult` struct (`action: 'continue' | 'paused' | 'synthesis_done'`) so the caller (the while loop) decides whether to continue or break. `steps: 0` is handled as "no tool calls allowed" — one text-only stream phase, tool calls ignored with a warn log. Step-cap hits produce a sentinel summary (reuses `cap_hit` kind so `CapHitSentinel.tsx` renders it without frontend changes; text distinguishes "Step limit reached" from "Tool budget exhausted"). Doom-loop check migrated from pre-recursion position to top of loop body — same predicate (`detectDoomLoop`), same threshold (3 identical calls), `break` instead of `return`. `step_start` parts are in the schema CHECK but not emitted as message_parts in v1.14 — writing to the assistant message before the stream phase creates a sequence-0 collision with `partsFromAssistantMessage`; a structured log line is emitted instead. Adversarial review caught the collision pre-deploy. 332/332 server tests passing; no frontend changes. Pairs with `v1.13.20-drop-legacy-cols` (parts is now the sole source of truth, and this batch's loop operates entirely through parts).
+
+## v1.13.20-drop-legacy-cols — 2026-05-23
+
+Final phase of the v1.13.0 strangler-fig migration. Removes the dual-write into `messages.tool_calls` / `messages.tool_results` JSON columns and drops the columns themselves; `message_parts` is now the only source of truth for tool-call and tool-result data. 10 dual-write sites stripped (5 in `tool-phase.ts`, 2 in `routes/skills.ts`, 2 in `routes/messages.ts`, 1 in `routes/chats.ts` fork-clone) — recon's grep-driven inventory caught 2 sites beyond the original v1.13.2 roadmap count. `messages_with_parts` view simplified to parts-only subselects (COALESCE fallbacks gone) and rewritten via `CREATE OR REPLACE VIEW` BEFORE the column DROP since Postgres rejects column-drop on view-referenced cols. Adversarial review caught a runtime bug the green test suite missed: `chats.ts:/api/chats/:id/discard_stale` had a `RETURNING ... tool_calls, tool_results, ...` clause referencing the dropped columns; would have crashed on every 60s-no-token-activity recovery in production. Fixed by switching to two-step UPDATE-then-SELECT-from-view so the response keeps the parts-synthesized fields. `Message` API type retains `tool_calls?` / `tool_results?` fields (override on the original v1.13.2 plan) — the view continues to populate them from parts, so the wire shape is unchanged and the frontend needs no updates. v1.12.1 cleanup block (`DROP CONSTRAINT messages_status_check`/`messages_role_check`) removed — those one-shots have done their work. `tool_cost_stats.test.ts` had a direct `INSERT INTO messages` touching the legacy columns that wasn't in the roadmap's inventory; rewritten to parts-table inserts and confirmed semantically faithful. 339/339 server tests passing including the 7 DB-integration tests (live-DB applied the schema migration and ran the parts-only view end-to-end). Pairs with `v1.13.0-ai-sdk-v6` (which introduced the dual-write) and `v1.13.1-B` (which moved the read path to `messages_with_parts`); umbrella `v1.13` tag ships on the same commit.
+
+## v1.13.19-html-artifact-panes — 2026-05-23
+
+Pane-based artifact viewer with on-request HTML support. Every assistant message gets an "Open in pane" icon button (`PanelRightOpen`, mobile 44px tap-target) in `MessageBubble`'s ActionRow; click opens the message in the workspace splitter as either a Markdown pane (Copy raw source + Download `.md`) or an HTML pane (Download `.html` only, no Copy). The HTML path triggers when the model emits a self-contained `<!DOCTYPE html>` or fenced ` ```html` artifact (opt-in only — `BOOCHAT.md` rule says Markdown is default at every length; HTML only on explicit user request like "render this as HTML"). Backend detection in `finalizeCompletion` (`error-handler.ts`) writes a new `message_parts.kind='html_artifact'` row with payload `{html_content, char_count, title}` (`<title>` → first `<h1>` → first 80 chars of inner text). Schema CHECK extended via the v1.13.13 drop-and-re-add pattern. 1MB cap is graceful — over-cap artifacts skip the part write and plain content lands; decision factored into a pure `decideHtmlArtifactWrite` helper so the warn-and-skip branch is unit-testable without mocking the full InferenceContext. Pane state is reference-only (`{chat_id, message_id, title}`) — content is fetched on mount, keeping `sessions.workspace_panes` jsonb small and avoiding 1MB blobs riding the `session_workspace_updated` WS frame. New `services/artifacts.ts` ships slug derivation (Markdown: first `#` heading → first 6 words; HTML: `<title>` → `<h1>` → inner text) and write helpers that realpath the artifacts directory after `mkdir` to close a symlink-escape gap (`assertArtifactsDirSafe`). `routes/artifacts.ts` exposes POST `/api/chats/:id/messages/:msg_id/artifacts/download?fmt=md|html` (writes to `<projectRoot>/.boocode/artifacts/<slug>-<ts>.<ext>`) plus GET `/api/projects/:project_id/artifacts/:filename` with `Content-Disposition: attachment`, `X-Content-Type-Options: nosniff`, and `Content-Security-Policy: sandbox` defense-in-depth on LLM-served HTML. iframe sandbox locks to `allow-scripts allow-clipboard-write allow-downloads` with no `allow-same-origin` and uses `srcDoc` (not `src`) for opaque-origin isolation. Frontend extracts `MarkdownRenderer.tsx` from `MessageBubble`'s inline `MarkdownBody` for reuse; `MarkdownArtifactPane.tsx` / `HtmlArtifactPane.tsx` render with loading + error states. 404-vs-real-error discrimination in `openInPane`: a real network/500 failure toasts and bails instead of silently masquerading as a Markdown pane. 31 new server unit tests (slug derivation, detection positive/negative, write helpers, symlink-escape, 1MB cap, real-symlink filesystem test); 332/332 server tests passing; `tsc -p apps/web/tsconfig.app.json --noEmit` clean; `pnpm -C apps/web build` green. Smoke deferred to first deploy.
+
+## v1.13.18-codecontext-file-path — 2026-05-22
+
+Fix: four codecontext wrappers (`get_file_analysis`, `get_symbol_info`, `get_dependencies`, `get_semantic_neighborhoods`) forwarded `file_path` to the sidecar unchanged, but the sidecar's index is keyed on absolute paths — every relative path from the model returned "File not found in graph" (three back-to-back failures in one chat at 17:56 UTC, ~48 s of wasted tool budget). New `resolveProjectPath` helper in `codecontext_client.ts:64-89` realpath-resolves the candidate, applies the same escape check as the existing `target_dir` resolver (matching the error template byte-for-byte except the field name), and falls through with the normalised absolute on ENOENT so the sidecar issues its own self-correctable "File not found" error. Wired into `callCodecontext` once at the args-spread site — all four wrappers benefit without per-wrapper edits. `.trim()` added to all four `file_path` Zod schemas to absorb trailing newlines from model output. Adversarial review caught a P2 escape-bypass: an absolute path with `..` (e.g. `<projectRoot>/../etc/passwd`) that ENOENTs at realpath would slip through the literal prefix-check, fixed by `resolve()`-normalising the absolute branch too. 9 new test cases in `codecontext_client.test.ts` (7 spec scenarios + symlink-out-of-root + absolute-with-`..` ENOENT) plus a 1-line update in `codecontext_tools.test.ts` asserting the new resolved-absolute contract. Pairs with `v1.13.17-cross-repo-reads` — both harden path traversal, but v1.13.18 stays inside the project root while v1.13.17 widens access outside it.
+
+## v1.13.17-cross-repo-reads — 2026-05-22
+
+On-demand read access to paths outside the session's primary project root. Closes the dead-end where `pathGuard` rejected every cross-repo read with no recovery path. New `request_read_access(path, reason)` tool emits an `ask_user_input`-style pause; user picks Allow/Deny via inline chips in `RequestReadAccessCard.tsx`; on Allow, the new `POST /api/chats/:id/grant_read_access` endpoint re-resolves the grant root and appends to `sessions.allowed_read_paths` (new `TEXT[]` column, default empty). Grant unit per design D1 = nearest registered `projects.path` ancestor → else nearest repo-shaped ancestor (`.git/` / `package.json` / `go.mod` / `Cargo.toml`) under `PROJECT_ROOT_WHITELIST` → else refuse without prompting. `pathGuard` extended with an optional `extraRoots` argument threaded from `session.allowed_read_paths` through `executeToolCall` to the four filesystem tools (view_file, list_dir, grep, find_files); `view_file` re-anchors the secret-guard check on `basename(real)` whenever the path resolved via a grant root so `.env` / `id_rsa*` deny still fires across grants. `grant_resolver.ts`'s ancestor walk checks the whitelist invariant on every iteration (not just final parent) so a symlinked input can't escape mid-walk. PATCH `/api/sessions/:id` exposes `allowed_read_paths` only for revocation: zod refines paths to absolute + no traversal markers, and a runtime subset guard (`findUnauthorizedAdditions`) rejects any entry not already present in the row, so a malicious `curl -X PATCH -d '{"allowed_read_paths":["/etc"]}'` 400s instead of bypassing the grant flow. Settings pane gains a per-session revoke list; archiving the session clears grants implicitly. 11 grant_resolver tests pin the symlink-escape-mid-walk guard (Sam's checkpoint-1 ask) and the nearest-project disambiguation; 8 path_guard tests cover extraRoots traversal; 8 sessions PATCH tests cover the subset guard including the `/etc` bypass attempt. Pairs with `v1.13.16-xml-parser` (model now both self-recovers from a wrong tool name AND from a refused path).
+
 ## v1.13.16-xml-parser — 2026-05-22

 Two-part fix for the model-emitted XML drift the v1.13.15 investigation surfaced. **Parser extension:** `xml-parser.ts` now recognizes the Anthropic `<invoke name="…"><parameter name="…">…</parameter></invoke>` shape alongside the existing Qwen/Hermes `<tool_call><function=…>…</function></tool_call>` shape. qwen3.6-35b-a3b-mxfp4 drifts to the Anthropic format when prompted as an Architect-style agent (Claude Code documentation in its pre-training corpus). Both formats route through the same synthetic-id `xml_call_${idx}` ToolCall path. The existing Qwen parser was tightened to tolerate whitespace around `=` (`<function = name>` shape) so a stray space doesn't get absorbed into the function name. **Unknown-tool recovery hint:** new `tool-suggestions.ts` exports `levenshtein()` + `suggestToolName()` + `formatUnknownToolError()`. When the dispatcher (`tool-phase.ts:executeToolCall`) receives an unknown tool name, the error returned to the model includes a "Did you mean: X?" hint based on Levenshtein distance ≤3 or substring match against `Object.keys(TOOLS_BY_NAME)`. Targets the qwen3.6 drift to `read_file` → suggest `view_file`. Test coverage in `xml-parser.test.ts` (46 tests, all green) covers both parsers, the partial-opener detector for both flavors, the unified extraction helper, and the new error formatter.
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -46,7 +46,7 @@ Tests: `pnpm -C apps/server test` runs the vitest suite. No test harness on `app
 - **Zod** for request validation and config parsing.

 Key services:
- **`services/inference/`** — Public surface re-exported via `inference/index.ts`; callers import from `./services/inference/index.js` explicitly (NodeNext doesn't honor directory-index resolution). Layout: `turn.ts` (runAssistantTurn / runInference / createInferenceRunner; exports `InferenceFrame`, `InferenceContext`, `TurnArgs`, `StreamResult`), `stream-phase.ts` (streamCompletion as a v1.13.1-A AI SDK adapter + executeStreamPhase), `provider.ts` (`upstreamModel(baseURL, modelId)` wrapping `createOpenAICompatible` against llama-swap), `tool-phase.ts` (executeToolPhase; value back-edges into turn.ts for the runAssistantTurn recursion — cycle safe because deref at call time, not module top-level), `sentinel-summaries.ts` (runCapHitSummary + runDoomLoopSummary + their sentinel inserters), `error-handler.ts` (handleAbortOrError, finalizeCompletion), `payload.ts` (buildMessagesPayload, loadContext, maybeFlagForCompaction, `OpenAiMessage`), `sentinels.ts` (`detectDoomLoop`, `DOOM_LOOP_THRESHOLD`, sentinel predicates), `budget.ts` (resolveToolBudget), `xml-parser.ts` (qwen3.6 XML tool-call fallback — KEEP, AI SDK doesn't handle inline-XML tool calls), `parts.ts` (v1.13.0 dual-write helpers: `partsFromAssistantMessage`, `partsFromToolMessage`, `insertParts`), `prune.ts` (v1.13.4 two-tier compaction; `selectPruneTargets` is the pure decision helper), `types.ts` (`StreamPhaseState`, `DB_FLUSH_INTERVAL_MS`). **`TurnArgs`** is the per-turn state envelope threaded through the `executeToolPhase → runAssistantTurn` recursion; reset in `runInference` at user-message boundary. Add new per-turn state to `TurnArgs`, not module-level closures.
+- **`services/inference/`** — Public surface re-exported via `inference/index.ts`; callers import from `./services/inference/index.js` explicitly (NodeNext doesn't honor directory-index resolution). Layout: `turn.ts` (runAssistantTurn / runInference / createInferenceRunner; exports `InferenceFrame`, `InferenceContext`, `TurnArgs`, `StreamResult`, `MAX_STEPS`), `stream-phase.ts` (streamCompletion as a v1.13.1-A AI SDK adapter + executeStreamPhase), `provider.ts` (`upstreamModel(baseURL, modelId)` wrapping `createOpenAICompatible` against llama-swap), `tool-phase.ts` (executeToolPhase → returns `ToolPhaseResult`; no longer recurses into runAssistantTurn — v1.14.0 converted the recursion to an explicit while loop in turn.ts), `sentinel-summaries.ts` (runCapHitSummary + runDoomLoopSummary + runStepCapSummary + their sentinel inserters), `error-handler.ts` (handleAbortOrError, finalizeCompletion), `payload.ts` (buildMessagesPayload, loadContext, maybeFlagForCompaction, `OpenAiMessage`), `sentinels.ts` (`detectDoomLoop`, `DOOM_LOOP_THRESHOLD`, sentinel predicates), `budget.ts` (resolveToolBudget), `xml-parser.ts` (qwen3.6 XML tool-call fallback — KEEP, AI SDK doesn't handle inline-XML tool calls), `parts.ts` (parts-table write helpers: `partsFromAssistantMessage`, `partsFromToolMessage`, `insertParts` — v1.13.20 made parts the sole source of truth), `prune.ts` (v1.13.4 two-tier compaction; `selectPruneTargets` is the pure decision helper), `types.ts` (`StreamPhaseState`, `DB_FLUSH_INTERVAL_MS`). **`TurnArgs`** is the per-turn state envelope populated from loop locals each iteration; reset in `runInference` at user-message boundary. The outer loop in `runAssistantTurn` (v1.14.0) runs `while (stepNumber < effectiveCap)` where `effectiveCap = Math.min(agent.steps ?? Infinity, MAX_STEPS=200)`. Per-agent `steps:` field in AGENTS.md frontmatter. `steps: 0` means text-only (no tool execution). Step-cap hit writes a `cap_hit` sentinel so `CapHitSentinel.tsx` renders it.
 - **AI SDK v6 streamCompletion adapter** (v1.13.1-A; `services/inference/stream-phase.ts`). `streamText` is the underlying call; the BooCode layer above (executeStreamPhase, finalize, dual-write) is shape-preserved via an adapter. Five gotchas the LSP/test suite won't catch:
  - **Abort signals are swallowed.** `streamText`'s `fullStream` iterator exits cleanly when `abortSignal` fires — no throw. Post-iteration `if (signal?.aborted) throw <AbortError>` is required; without it the row finalizes as `complete` instead of `cancelled`. Comment in stream-phase.ts pins this; don't refactor it away.
  - **Usage lands only at stream end** via `await result.usage` (`inputTokens` / `outputTokens` v6 names → mapped to `promptTokens` / `completionTokens` for the existing onUsage callback). Mid-stream live tok/s is gone vs v1.12.2; ChatThroughput shows a single value at stream end.
@@ -63,12 +63,20 @@ Key services:
 - **`services/compaction.ts`** + **`services/model-context.ts`** — v1.11.0 anchored rolling summary (single `summary=true` assistant row per chat, supersedes itself on each compaction). Triggered when `chats.needs_compaction` is set after an inference turn exceeds `usable(ctx_max) = floor(0.85 × ctx_max)` (v1.13.9 opencode-pattern early trigger; was `ctx_max - 20k` pre-v1.13.9, which gave only 7.6% headroom at 262k and 0 budget for ≤20k contexts). **`ctx_max` comes from `model-context.getModelContext()` which fetches `${LLAMA_SWAP_URL}/upstream/<model>/props`** — NOT from `parsed.timings.n_ctx` (the stream completion's `timings` doesn't carry n_ctx; that read was dead code until v1.11.3 ripped it out). First inferences after a boocode boot may have `ctx_max=NULL` if llama-swap hasn't loaded the model yet; negative cache TTL is 60s, recovers on next turn. v1.13.6: `buildHeadPayload` embeds `reasoning_parts` as a `<reasoning>...</reasoning>` prose prefix on the assistant `content` (OpenAI wire shape has no structured reasoning field; the summarizer reads text). Standalone tag when content is empty (tool-call-only turn). `buildHeadPayload` + `OpenAiMessage` exported for test access — keep them exported.
 - **`services/system-prompt.ts`** — `buildSystemPrompt` is the string-returning shim; `buildSystemPromptWithFingerprint` is the canonical impl returning `{prompt, fingerprint, drift}`. v1.13.8 instrumentation: SHA-256 of the assembled prefix is logged per `buildMessagesPayload` call (msg `prefix-fingerprint`, level=info); a `Map<sessionId, lastHash>` observer fires `prefix-drift` (level=warn) on hash change with a field-level `changed_inputs` diff. Smoke proved the prefix is byte-stable across turns in steady-state — the originally-planned `system_prompt_cache` DB table was dropped as redundant against the v1.12.0 input-layer mtime caches (BOOCHAT.md here + AGENTS.md global+per-project in `agents.ts:safeStat`).
 - **`services/inference/budget.ts`** — tool-call budgets: `BUDGET_READ_ONLY = 30`, `BUDGET_NON_READ_ONLY = 10` (forward-looking; no write tools yet), `BUDGET_NO_AGENT = 30` (v1.13.7; was 15 — every tool in `ALL_TOOLS` is read-only today, so no-agent mode shares the read-only-agent cap). Per-agent `max_tool_calls` from AGENTS.md frontmatter overrides.
- **`messages_with_parts` view** (v1.13.1-B; `schema.sql`). Read sites that need `tool_calls` / `tool_results` / `reasoning_parts` SELECT from this view, NOT `messages` directly. `COALESCE`s parts-table rows over the legacy JSON columns, so pre-v1.13.0 history still resolves. Writes still target `messages`; the v1.13.0 dual-write into `message_parts` keeps both halves in sync. New payload-assembly code must use the view — calling `messages.tool_calls` directly will miss anything written post-v1.13.1-B if the JSON column ever drifts (and dual-write makes that easy to miss). Shapes: `tool_calls jsonb[]`, `tool_results jsonb` single object, `reasoning_parts jsonb[]` of `{text}`.
+- **`messages_with_parts` view** (v1.13.1-B; `schema.sql`). Read sites that need `tool_calls` / `tool_results` / `reasoning_parts` SELECT from this view, NOT `messages` directly. v1.13.20 dropped the legacy `messages.tool_calls` / `messages.tool_results` JSON columns; the view now reads parts-only subselects. Writes target `message_parts` exclusively via `insertParts` (or via the helpers `partsFromAssistantMessage` / `partsFromToolMessage`). The `Message` wire type still carries `tool_calls?` / `tool_results?` because the view synthesizes them from parts — frontend reads are unchanged. Shapes: `tool_calls jsonb[]`, `tool_results jsonb` single object, `reasoning_parts jsonb[]` of `{text}`. If you ever need to UPDATE a message and return its full Message shape, do a two-step UPDATE returning `id` followed by SELECT from the view — RETURNING off the bare `messages` table no longer carries the tool fields.
 - **`services/file_ops.ts`** — Shared file operation implementations used by both inference tools and HTTP routes.
 - **`services/auto_name.ts`** — Non-streaming LLM call to generate 4-word session titles after first assistant reply.

 Route registration: all routes registered in `index.ts` via `register*Routes(app, sql, ...)` functions. Routes are in `routes/*.ts`.

+### BooCoder (`apps/coder/src/`)
+
+- Write-capable coding agent. Separate Fastify server at port 9502, same docker network (`boocode_net`).
+- **Workspace dependency on `@boocode/server`**: imports `createInferenceRunner`, `createBroker`, `ALL_TOOLS`, `appendMcpTools` from the server's compiled `dist/`. apps/server's `package.json` has an `exports` map with `types` conditions for NodeNext resolution. apps/server must build FIRST (Dockerfile builds server → coder).
+- `apps/server/tsconfig.json` has `declaration: true` so `.d.ts` files exist for workspace consumers.
+- Write tools (`edit_file`, `create_file`, `delete_file`, `apply_pending`, `rewind`) queue in `pending_changes` table. Nothing hits disk until `apply_pending` is called. `write_guard.ts` validates paths (resolve + prefix-check, no realpath since files may not exist for creates).
+- Frontend: NOT a separate SPA. BooCoder is a `'coder'` pane type within BooChat's SPA (`apps/web/`). `CoderPane.tsx` in `apps/web/src/components/panes/`. API requests go through `/api/coder/*` proxy (Vite dev + Fastify production) which rewrites to `http://boocoder:3000/api/*`. WS connects directly to `:9502`.
+
 ### Frontend (`apps/web/src/`)

 - **React 18** + React Router v6 + **Tailwind v4** + shadcn/radix-ui primitives.
@@ -105,19 +113,23 @@ Sessions hold 1–5 panes (chat / empty / placeholder terminal+agent). v1.12.1 m

 ## Database

-PostgreSQL 16. Tables: `projects`, `sessions`, `chats`, `messages`, `settings`, `message_parts` (v1.13.0). Views: `messages_with_parts` (v1.13.1-B parts-merge read path), `tool_cost_stats` (v1.13.10 per-tool 100-call rolling window). (`session_panes` was dropped in v1.12.1; workspace pane state lives in `sessions.workspace_panes jsonb`.) Schema applied idempotently on startup via `applySchema()`. Use `clock_timestamp()` (not `NOW()`) inside transactions. CHECK constraints in place: `projects_status_chk` ('open'|'archived'), `sessions_status_chk` (same), `chats_status_chk` (same), `messages_role_chk`, `messages_status_chk` — keep in sync with the `*_STATUSES` const arrays in `apps/server/src/types/api.ts`. The older anonymous `messages_status_check` (without 'cancelled') and `messages_role_check` (without 'system') were dropped in v1.12.1; only the `_chk` variants remain.
+PostgreSQL 16. Database name: `boochat` (renamed from `boocode` in v2.0.0-alpha; Docker service name stays `boocode_db`). Tables: `projects`, `sessions`, `chats`, `messages`, `settings`, `message_parts` (v1.13.0), `pending_changes` (v2.0.0), `tasks` (v2.0.0), `available_agents` (v2.0.0). Views: `messages_with_parts` (v1.13.1-B parts-merge read path), `tool_cost_stats` (v1.13.10 per-tool 100-call rolling window), `human_inbox` (v2.0.0 — tasks WHERE state IN blocked/failed). (`session_panes` was dropped in v1.12.1; workspace pane state lives in `sessions.workspace_panes jsonb`.) Schema applied idempotently on startup via `applySchema()`. Use `clock_timestamp()` (not `NOW()`) inside transactions. CHECK constraints in place: `projects_status_chk` ('open'|'archived'), `sessions_status_chk` (same), `chats_status_chk` (same), `messages_role_chk`, `messages_status_chk` — keep in sync with the `*_STATUSES` const arrays in `apps/server/src/types/api.ts`. The older anonymous `messages_status_check` (without 'cancelled') and `messages_role_check` (without 'system') were dropped in v1.12.1; only the `_chk` variants remain.

 Schema CHECK migration order when renaming allowed values: (1) `ALTER TABLE ... DROP CONSTRAINT IF EXISTS <system_name>` (inline `CREATE TABLE` checks get `<table>_<column>_check`), (2) `UPDATE` rows to new values, (3) wrap new constraint ADD in `DO $$ ... pg_constraint` guard — that block is the only way to get `ADD CONSTRAINT IF NOT EXISTS`.


 ## Environment

-Required: `DATABASE_URL`, `LLAMA_SWAP_URL`. Optional: `PORT` (3000), `HOST` (0.0.0.0), `PROJECT_ROOT_WHITELIST` (/opt, read-only scope for add-existing path resolution), `BOOTSTRAP_ROOT` (/opt/projects, writable scope for create-new-project bootstrap mkdir target — host must `mkdir -p /opt/projects` before container start), `DEFAULT_MODEL`, `LOG_LEVEL`, `SEARXNG_URL` (default `http://100.114.205.53:8888` — internal Tailscale Fathom; the public `search.indifferentketchup.com` is behind Authelia and unusable from server context), `BOOCODE_TOOLS` (`core` | `standard` | `all`, default `all`; v1.13.15-tools tier filter — ceiling, never expands an agent's whitelist).
+Required: `DATABASE_URL`, `LLAMA_SWAP_URL`. Optional: `PORT` (3000), `HOST` (0.0.0.0), `PROJECT_ROOT_WHITELIST` (/opt, read-only scope for add-existing path resolution), `BOOTSTRAP_ROOT` (/opt/projects, writable scope for create-new-project bootstrap mkdir target — host must `mkdir -p /opt/projects` before container start), `DEFAULT_MODEL`, `LOG_LEVEL`, `SEARXNG_URL` (default `http://100.114.205.53:8888` — internal Tailscale Fathom; the public `search.indifferentketchup.com` is behind Authelia and unusable from server context), `BOOCODE_TOOLS` (`core` | `standard` | `all`, default `all`; v1.13.15-tools tier filter — ceiling, never expands an agent's whitelist), `MCP_CONFIG_PATH` (optional; default `/data/mcp.json` — JSON config for MCP servers matching opencode's `mcpServers` shape; file missing = no MCP).
+
+BooCoder at port 9502: `curl http://100.114.205.53:9502/api/health`. Same Tailscale IP binding as BooChat. Health reports tool count: `{"ok":true,"db":true,"tools":30}`.

 ## Workflow

 - Sam reviews all diffs and commits manually. Do not commit unless explicitly asked.
 - Per-batch docs live under `openspec/changes/<slug>/{proposal,tasks,design}.md`. Already-shipped batches are snapshots in `openspec/changes/archived/`. New batches follow the proposal+tasks shape; see `openspec/README.md` for the convention.
+- Tag naming: `vMAJOR.MINOR.PATCH-slug` (e.g. `v1.13.13-ws-publish`). Monotonic per minor — the slug describes the batch's content so the tag name alone is enough to recall what shipped. No letter suffixes (`-a`/`-b`), no pseudo-ranges (`v1.11.x`), no slug-only sub-versions sharing a number (`v1.13.15-tools` + `-openspec` + `-agentlint` — split into sequential patches instead).
+- `CHANGELOG.md` is the per-tag release log, most-recent on top. When a new tag is created, add a `## <tag> — <YYYY-MM-DD>` section with a 3–6 sentence paragraph summarizing what shipped, drawn from the commit body. Cross-reference other tags by name when the batch builds on, fixes, or pairs with prior work (e.g. "pairs with `v1.13.12-ws-schemas`", "fixed in `v1.13.5-stability-bundle`"). No nested bullets — one paragraph.
 - Deploy: `cd /opt/boocode && docker compose up --build -d` (or `docker compose build --no-cache boocode && docker compose up -d` if you suspect a layer-cache issue).
 - Git push to Gitea: `GIT_SSH_COMMAND="ssh -i /opt/boocode/secrets/boocode_gitea -o IdentitiesOnly=yes" git push origin <branch>`. The default agent identity is rejected; the in-repo deploy key (`secrets/`, gitignored) is the working one. Transient `Connection reset by peer` retries cleanly after `sleep 5`.
 - Don't accumulate `.bak-*` files. Clean them up in the same batch or immediately after merge.
@@ -132,6 +144,8 @@ Required: `DATABASE_URL`, `LLAMA_SWAP_URL`. Optional: `PORT` (3000), `HOST` (0.0
 - `/opt/boolab` hosts a working sibling BooCode terminal at `boocode.indifferentketchup.com`. Useful for visual side-by-side comparison on the same iPhone when debugging booterm rendering. Boolab uses Tailwind v3 (`@tailwind base`); boocode uses v4 — many subtle build differences. Don't assume parity.
 - booterm SSHs to the host as `samkintop@100.114.205.53` (the Tailscale IP). The hostname `ubuntu-homelab` (shown in the bash prompt after login) does NOT resolve from inside the container — only the host's `/etc/hosts` knows it. Override via `BOOTERM_SSH_HOST` / `BOOTERM_SSH_USER` env vars in docker-compose if you ever move the shell to a different machine.
 - codecontext sidecar lives at `/opt/boocode/codecontext/`. Sidecar HTTP API at `http://codecontext:8080/v1/<tool_name>` over the `boocode_net` bridge (no host port). BooCode wrappers in `apps/server/src/services/tools/codecontext/`. The `.codecontextignore.template` documents recommended ignore patterns; users copy and adapt to project root manually.
+- codecontext fork at `/opt/forks/codecontext/` — separate git repo (branch `boocode-ts`), pushed via the same boocode_gitea SSH key to `indifferentketchup/codecontext`. Build: `go build ./...`. Test: `go test ./...`. Docker rebuild: `docker compose build --no-cache codecontext`.
+- Go binary: `/snap/go/current/bin/go` (not on PATH by default). Use `export PATH=$PATH:/snap/go/current/bin` or full path for Go commands.
 - `os/exec` child supervisors must explicitly call `child.Wait()` in a goroutine and `os.Exit` on child death. `Signal(0)` returns nil on zombies and is NOT a liveness check. Without `Wait()`, docker's `restart: unless-stopped` policy never fires because the parent stays alive. The `codecontext/shim.go` implementation is the reference pattern.

 ## Conventions
@@ -154,3 +168,5 @@ Required: `DATABASE_URL`, `LLAMA_SWAP_URL`. Optional: `PORT` (3000), `HOST` (0.0
 - Tool-name whitelists must derive from `ALL_TOOLS` in `services/tools.ts`, never hardcoded. `services/agents.ts` `ALL_TOOL_NAMES` had this drift class until v1.12 — same pattern applies to any future tool-aware code.
 - Agent registry lives at `data/AGENTS.md` (global, bind-mounted at `/data/AGENTS.md`). No per-project `AGENTS.md` in this repo — removed in v1.12 to eliminate the two-files-must-stay-in-sync drift. The `getAgentsForProject` per-project override mechanism remains for *other* projects.
 - MCP stdio transport uses newline-delimited JSON (NDJSON), NOT LSP-style `Content-Length` headers. The `codecontext/shim.go` framing implementation is the reference; per the MCP spec (modelcontextprotocol.io/specification/server/transports).
+- **Workspace dependency pattern** (`apps/coder` → `@boocode/server`): the consuming package adds `"@boocode/server": "workspace:*"` in `package.json`. The provider's `package.json` needs `exports` with `types` + `default` conditions per subpath: `"./inference": { "types": "./dist/.../index.d.ts", "default": "./dist/.../index.js" }`. Without the `types` condition, NodeNext resolution can't find `.d.ts` files and tsc fails with "Cannot find module" in the consumer.
+- **Docker build order for workspace deps**: the Dockerfile must `COPY` + `RUN pnpm build` the provider app BEFORE the consumer app. `apps/coder/Dockerfile` builds `apps/server` first, then `apps/coder`.
--- a/apps/coder/Dockerfile
+++ b/apps/coder/Dockerfile
@@ -0,0 +1,35 @@
+# syntax=docker/dockerfile:1.7
+
+FROM node:20-alpine AS builder
+RUN corepack enable
+WORKDIR /build
+
+COPY package.json pnpm-workspace.yaml pnpm-lock.yaml tsconfig.base.json ./
+COPY apps/server/package.json ./apps/server/
+COPY apps/coder/package.json ./apps/coder/
+COPY apps/coder/web/package.json ./apps/coder/web/
+
+RUN pnpm install --frozen-lockfile
+
+# Build server first (coder depends on it via workspace dep for types + inference)
+COPY apps/server ./apps/server
+RUN pnpm -C apps/server build
+
+COPY apps/coder ./apps/coder
+RUN pnpm -C apps/coder/web build
+RUN pnpm -C apps/coder build
+
+RUN pnpm deploy --filter=@boocode/coder --prod --legacy /out/coder
+
+
+FROM node:20-bookworm-slim AS runtime
+RUN apt-get update && apt-get install -y --no-install-recommends ripgrep git openssh-client && rm -rf /var/lib/apt/lists/*
+WORKDIR /app
+
+COPY --from=builder /out/coder ./
+COPY --from=builder /build/apps/coder/web/dist ./web
+
+ENV NODE_ENV=production
+EXPOSE 3000
+
+CMD ["node", "dist/index.js"]
--- a/apps/coder/package.json
+++ b/apps/coder/package.json
@@ -0,0 +1,33 @@
+{
+  "name": "@boocode/coder",
+  "version": "2.0.0",
+  "private": true,
+  "type": "module",
+  "main": "dist/index.js",
+  "scripts": {
+    "dev": "tsx watch src/index.ts",
+    "build": "tsc && node -e \"import('node:fs').then(fs=>fs.copyFileSync('src/schema.sql','dist/schema.sql'))\"",
+    "start": "node dist/index.js",
+    "cli": "tsx src/cli.ts",
+    "typecheck": "tsc --noEmit",
+    "test": "vitest run"
+  },
+  "dependencies": {
+    "@agentclientprotocol/sdk": "^0.22.1",
+    "@boocode/server": "workspace:*",
+    "@fastify/static": "^7.0.4",
+    "@fastify/websocket": "^10.0.1",
+    "@modelcontextprotocol/sdk": "^1.29.0",
+    "fastify": "^4.28.1",
+    "postgres": "^3.4.4",
+    "ws": "^8.18.0",
+    "zod": "^3.23.8"
+  },
+  "devDependencies": {
+    "@types/node": "^20.14.10",
+    "@types/ws": "^8.5.10",
+    "tsx": "^4.16.2",
+    "typescript": "^5.5.0",
+    "vitest": "^3.0.0"
+  }
+}
--- a/apps/coder/src/cli.ts
+++ b/apps/coder/src/cli.ts
@@ -0,0 +1,249 @@
+#!/usr/bin/env node
+/**
+ * BooCoder CLI client.
+ *
+ * Usage:
+ *   boocode run "task description" [--agent opencode] [--model claude-opus-4-7] [--project <id>]
+ *   boocode ls [--state pending|running|completed|failed]
+ *   boocode attach <task-id>
+ *   boocode send <task-id> "message"
+ */
+import { WebSocket } from 'ws';
+
+const BASE_URL = process.env.BOOCODER_URL ?? 'http://100.114.205.53:9502';
+
+// ─── Arg parsing ─────────────────────────────────────────────────────────────
+
+function getFlag(args: string[], name: string): string | undefined {
+  const idx = args.indexOf(name);
+  if (idx === -1 || idx + 1 >= args.length) return undefined;
+  return args[idx + 1];
+}
+
+function hasFlag(args: string[], name: string): boolean {
+  return args.includes(name);
+}
+
+// ─── HTTP helpers ────────────────────────────────────────────────────────────
+
+async function api(method: string, path: string, body?: unknown): Promise<unknown> {
+  const url = `${BASE_URL}${path}`;
+  const res = await fetch(url, {
+    method,
+    headers: body ? { 'Content-Type': 'application/json' } : undefined,
+    body: body ? JSON.stringify(body) : undefined,
+  });
+  if (!res.ok) {
+    const text = await res.text().catch(() => '');
+    throw new Error(`${method} ${path} → ${res.status}: ${text}`);
+  }
+  return res.json();
+}
+
+// ─── WS streaming ────────────────────────────────────────────────────────────
+
+function streamSession(sessionId: string): void {
+  const wsUrl = BASE_URL.replace(/^http/, 'ws') + `/api/ws/sessions/${sessionId}`;
+  const ws = new WebSocket(wsUrl);
+
+  ws.on('message', (data) => {
+    try {
+      const frame = JSON.parse(data.toString()) as { type: string; content?: string; name?: string; arguments?: string };
+      if (frame.type === 'delta' && frame.content) {
+        process.stdout.write(frame.content);
+      } else if (frame.type === 'tool_call') {
+        process.stdout.write(`\n[tool: ${frame.name ?? '?'}(${(frame.arguments ?? '').slice(0, 80)})]\n`);
+      } else if (frame.type === 'tool_result') {
+        process.stdout.write(`[tool_result]\n`);
+      } else if (frame.type === 'status' || frame.type === 'chat_status') {
+        // Silent
+      }
+    } catch {
+      // Non-JSON frame, ignore
+    }
+  });
+
+  ws.on('error', (err) => {
+    process.stderr.write(`WS error: ${err.message}\n`);
+  });
+
+  ws.on('close', () => {
+    process.stdout.write('\n');
+    process.exit(0);
+  });
+
+  process.on('SIGINT', () => {
+    ws.close();
+    process.exit(0);
+  });
+}
+
+// ─── Commands ────────────────────────────────────────────────────────────────
+
+async function cmdRun(args: string[]): Promise<void> {
+  const input = args.find((a) => !a.startsWith('--'));
+  if (!input) {
+    process.stderr.write('Usage: boocode run "task description" [--agent X] [--model X] [--project X]\n');
+    process.exit(1);
+  }
+
+  const agent = getFlag(args, '--agent');
+  const model = getFlag(args, '--model');
+  const project_id = getFlag(args, '--project');
+
+  if (!project_id) {
+    process.stderr.write('Error: --project <uuid> is required\n');
+    process.exit(1);
+  }
+
+  const result = (await api('POST', '/api/tasks', {
+    project_id,
+    input,
+    ...(agent && { agent }),
+    ...(model && { model }),
+  })) as { id: string; state: string };
+
+  process.stdout.write(`Task created: ${result.id} (state: ${result.state})\n`);
+
+  // Poll until task has session_id, then stream; or poll until terminal state
+  const POLL_MS = 2000;
+  for (;;) {
+    await sleep(POLL_MS);
+    const task = (await api('GET', `/api/tasks/${result.id}`)) as {
+      id: string; state: string; session_id?: string; output_summary?: string;
+    };
+
+    if (task.session_id) {
+      process.stdout.write(`Streaming session ${task.session_id}...\n`);
+      streamSession(task.session_id);
+      return; // streamSession handles exit
+    }
+
+    if (task.state === 'completed') {
+      process.stdout.write(`\nCompleted: ${task.output_summary ?? '(no summary)'}\n`);
+      return;
+    }
+    if (task.state === 'failed') {
+      process.stderr.write(`\nFailed: ${task.output_summary ?? '(no summary)'}\n`);
+      process.exit(1);
+    }
+    if (task.state === 'cancelled') {
+      process.stderr.write(`\nCancelled.\n`);
+      process.exit(1);
+    }
+  }
+}
+
+async function cmdLs(args: string[]): Promise<void> {
+  const state = getFlag(args, '--state');
+  const query = state ? `?state=${state}` : '';
+  const tasks = (await api('GET', `/api/tasks${query}`)) as Array<{
+    id: string; state: string; agent: string | null; input: string; created_at: string;
+  }>;
+
+  if (tasks.length === 0) {
+    process.stdout.write('No tasks.\n');
+    return;
+  }
+
+  // Table header
+  process.stdout.write(
+    pad('ID', 38) + pad('STATE', 12) + pad('AGENT', 14) + pad('INPUT', 52) + 'CREATED\n',
+  );
+  process.stdout.write('-'.repeat(120) + '\n');
+
+  for (const t of tasks) {
+    process.stdout.write(
+      pad(t.id, 38) +
+      pad(t.state, 12) +
+      pad(t.agent ?? '-', 14) +
+      pad(t.input.slice(0, 50), 52) +
+      (t.created_at?.slice(0, 19) ?? '') + '\n',
+    );
+  }
+}
+
+async function cmdAttach(args: string[]): Promise<void> {
+  const taskId = args[0];
+  if (!taskId) {
+    process.stderr.write('Usage: boocode attach <task-id>\n');
+    process.exit(1);
+  }
+
+  const task = (await api('GET', `/api/tasks/${taskId}`)) as { session_id?: string };
+  if (!task.session_id) {
+    process.stderr.write('Task has no session yet (still pending?).\n');
+    process.exit(1);
+  }
+
+  streamSession(task.session_id);
+}
+
+async function cmdSend(args: string[]): Promise<void> {
+  const taskId = args[0];
+  const message = args[1];
+  if (!taskId || !message) {
+    process.stderr.write('Usage: boocode send <task-id> "message"\n');
+    process.exit(1);
+  }
+
+  const task = (await api('GET', `/api/tasks/${taskId}`)) as { session_id?: string };
+  if (!task.session_id) {
+    process.stderr.write('Task has no session yet.\n');
+    process.exit(1);
+  }
+
+  // Find active chat
+  const sessionId = task.session_id;
+  // POST message to the session's chat (the messages route expects session_id in path)
+  await api('POST', `/api/sessions/${sessionId}/messages`, { content: message });
+
+  // Then attach to stream the response
+  streamSession(sessionId);
+}
+
+// ─── Utils ───────────────────────────────────────────────────────────────────
+
+function pad(s: string, width: number): string {
+  return s.length >= width ? s.slice(0, width) : s + ' '.repeat(width - s.length);
+}
+
+function sleep(ms: number): Promise<void> {
+  return new Promise((resolve) => setTimeout(resolve, ms));
+}
+
+// ─── Main ────────────────────────────────────────────────────────────────────
+
+const [cmd, ...rest] = process.argv.slice(2);
+
+switch (cmd) {
+  case 'run':
+    cmdRun(rest).catch(fatal);
+    break;
+  case 'ls':
+    cmdLs(rest).catch(fatal);
+    break;
+  case 'attach':
+    cmdAttach(rest).catch(fatal);
+    break;
+  case 'send':
+    cmdSend(rest).catch(fatal);
+    break;
+  default:
+    process.stdout.write(
+      'BooCoder CLI\n\n' +
+      'Commands:\n' +
+      '  run "task"  [--agent X] [--model X] [--project <id>]   Create and stream a task\n' +
+      '  ls          [--state pending|running|completed|failed]   List tasks\n' +
+      '  attach      <task-id>                                    Stream a running task\n' +
+      '  send        <task-id> "message"                          Send input to a task\n' +
+      '\n' +
+      `Base URL: ${BASE_URL} (set BOOCODER_URL to override)\n`,
+    );
+    if (cmd && cmd !== '--help' && cmd !== '-h') process.exit(1);
+}
+
+function fatal(err: unknown): void {
+  process.stderr.write(`Error: ${err instanceof Error ? err.message : String(err)}\n`);
+  process.exit(1);
+}
--- a/apps/coder/src/config.ts
+++ b/apps/coder/src/config.ts
@@ -0,0 +1,47 @@
+import { z } from 'zod';
+
+// BooCoder's config is a superset of the server's Config type so it can be
+// passed directly into the inference runner's InferenceContext. Fields the
+// inference loop reads: LLAMA_SWAP_URL, PROJECT_ROOT_WHITELIST. The rest
+// default to values that satisfy the server's Zod schema without BooCoder
+// needing to supply them in its environment.
+const ConfigSchema = z.object({
+  NODE_ENV: z.enum(['development', 'production', 'test']).default('development'),
+  PORT: z.coerce.number().int().positive().default(3000),
+  HOST: z.string().default('0.0.0.0'),
+  DATABASE_URL: z.string().url(),
+  LLAMA_SWAP_URL: z.string().url(),
+  PROJECT_ROOT_WHITELIST: z.string().default('/opt'),
+  BOOTSTRAP_ROOT: z.string().default('/opt/projects'),
+  DEFAULT_MODEL: z.string().default('qwen3.6-35b-a3b-mxfp4'),
+  LOG_LEVEL: z.string().default('info'),
+  CONTAINER_GUIDANCE_FILE: z.string().optional(),
+  // Fields needed to satisfy the server's Config type but unused by BooCoder:
+  SEARXNG_URL: z.string().url().default('http://100.114.205.53:8888'),
+  GITEA_BASE_URL: z.string().url().default('https://git.indifferentketchup.com'),
+  GITEA_USER: z.string().default('indifferentketchup'),
+  GITEA_TOKEN: z.string().optional(),
+  GITEA_SSH_HOST: z.string().default('100.114.205.53:2222'),
+  MCP_CONFIG_PATH: z.string().optional(),
+  // v2.0.5: cheaper model for titles, summaries, labeling.
+  FAST_MODEL: z.string().optional(),
+  // SSH access to the host for external agent dispatch (Phase 5)
+  BOOCODER_SSH_HOST: z.string().default('100.114.205.53'),
+  BOOCODER_SSH_USER: z.string().default('samkintop'),
+});
+
+export type Config = z.infer<typeof ConfigSchema>;
+
+let cached: Config | null = null;
+
+export function loadConfig(): Config {
+  if (cached) return cached;
+  const parsed = ConfigSchema.safeParse(process.env);
+  if (!parsed.success) {
+    console.error('Invalid environment configuration:');
+    console.error(parsed.error.flatten().fieldErrors);
+    process.exit(1);
+  }
+  cached = parsed.data;
+  return cached;
+}
--- a/apps/coder/src/db.ts
+++ b/apps/coder/src/db.ts
@@ -0,0 +1,45 @@
+import postgres from 'postgres';
+import { readFile } from 'node:fs/promises';
+import { fileURLToPath } from 'node:url';
+import { dirname, resolve } from 'node:path';
+import type { Config } from './config.js';
+
+const __filename = fileURLToPath(import.meta.url);
+const __dirname = dirname(__filename);
+
+export type Sql = ReturnType<typeof postgres>;
+
+let sqlInstance: Sql | null = null;
+
+export function getSql(config: Config): Sql {
+  if (sqlInstance) return sqlInstance;
+  sqlInstance = postgres(config.DATABASE_URL, {
+    max: 10,
+    idle_timeout: 30,
+    connect_timeout: 10,
+    onnotice: () => {},
+  });
+  return sqlInstance;
+}
+
+export async function applySchema(sql: Sql): Promise<void> {
+  const schemaPath = resolve(__dirname, 'schema.sql');
+  const ddl = await readFile(schemaPath, 'utf8');
+  await sql.unsafe(ddl);
+}
+
+export async function pingDb(sql: Sql): Promise<boolean> {
+  try {
+    await sql`SELECT 1`;
+    return true;
+  } catch {
+    return false;
+  }
+}
+
+export async function closeDb(): Promise<void> {
+  if (sqlInstance) {
+    await sqlInstance.end({ timeout: 5 });
+    sqlInstance = null;
+  }
+}
--- a/apps/coder/src/index.ts
+++ b/apps/coder/src/index.ts
@@ -0,0 +1,189 @@
+import { resolve, dirname } from 'node:path';
+import { fileURLToPath } from 'node:url';
+import { existsSync } from 'node:fs';
+import Fastify from 'fastify';
+import fastifyWebsocket from '@fastify/websocket';
+import fastifyStatic from '@fastify/static';
+
+const __filename = fileURLToPath(import.meta.url);
+const __dirname = dirname(__filename);
+import { loadConfig } from './config.js';
+import { getSql, applySchema, pingDb, closeDb } from './db.js';
+import { startMcpServer } from './services/mcp-server.js';
+// v2.0.0 Phase 2B: workspace dependency on @boocode/server — reuse the
+// inference loop, broker, and tool registry without duplication.
+import { createInferenceRunner } from '@boocode/server/inference';
+import { createBroker } from '@boocode/server/broker';
+import { appendMcpTools, ALL_TOOLS } from '@boocode/server/tools';
+import type { Config as ServerConfig } from '@boocode/server/config';
+import type { WsFrame } from '@boocode/server/ws-frames';
+// v2.0.0 Phase 2C: write tools + adapter for BooChat ToolDef compatibility.
+import { WRITE_TOOLS } from './services/tools/index.js';
+import { adaptWriteTool } from './services/tools/adapter.js';
+import { setInferenceContext, clearInferenceContext } from './services/tools/inference_context.js';
+// Routes
+import { registerMessageRoutes } from './routes/messages.js';
+import { registerPendingRoutes } from './routes/pending.js';
+import { registerTaskRoutes } from './routes/tasks.js';
+import { registerInboxRoutes } from './routes/inbox.js';
+import { registerStatsRoutes } from './routes/stats.js';
+import { registerArenaRoutes } from './routes/arena.js';
+import { registerWebSocket } from './routes/ws.js';
+// Phase 4: dispatcher + agent probe
+import { createDispatcher } from './services/dispatcher.js';
+import { probeAgents } from './services/agent-probe.js';
+
+async function main() {
+  // MCP mode: stdio transport, no HTTP server
+  if (process.argv.includes('--mcp')) {
+    const config = loadConfig();
+    const sql = getSql(config);
+    await applySchema(sql);
+    await startMcpServer(sql);
+    return;
+  }
+
+  const config = loadConfig();
+
+  const app = Fastify({
+    logger: { level: config.LOG_LEVEL },
+  });
+
+  // Allow empty JSON bodies (same pattern as apps/server).
+  app.removeContentTypeParser(['application/json']);
+  app.addContentTypeParser('application/json', { parseAs: 'string' }, (_req, body, done) => {
+    const str = (body as string) ?? '';
+    if (str.trim().length === 0) {
+      done(null, {});
+      return;
+    }
+    try {
+      done(null, JSON.parse(str));
+    } catch (err) {
+      done(err as Error, undefined);
+    }
+  });
+
+  const sql = getSql(config);
+  await applySchema(sql);
+  app.log.info('database schema applied');
+
+  // Broker: in-memory pub/sub for session + user channel streaming.
+  const broker = createBroker(app.log);
+
+  // --- Tool registry extension ---
+  // Append BooCoder write tools (adapted to BooChat's ToolDef interface) to
+  // the shared ALL_TOOLS registry. appendMcpTools re-sorts and rebuilds
+  // TOOLS_BY_NAME so tool-phase.ts dispatch sees the full set.
+  const adaptedWriteTools = WRITE_TOOLS.map((t) => adaptWriteTool(t));
+  appendMcpTools(adaptedWriteTools);
+  app.log.info(`tool registry: ${ALL_TOOLS.length} tools loaded (${WRITE_TOOLS.length} write tools)`);
+
+  // Inference runner: same engine as BooChat, uses ALL_TOOLS (which includes
+  // the appended write tools) for tool dispatch.
+  const inference = createInferenceRunner(
+    {
+      sql,
+      config: config as unknown as ServerConfig,
+      log: app.log,
+      publish: (sessionId, frame) => {
+        broker.publishFrame(sessionId, frame as unknown as WsFrame);
+      },
+      broker,
+    },
+    (user, frame) => {
+      broker.publishUserFrame(user, frame as unknown as WsFrame);
+    }
+  );
+
+  // Wrap the inference runner to set/clear the write-tool context around each run.
+  // The inference runner calls enqueue() which fires asynchronously — we hook
+  // into the enqueue to set context before the run starts.
+  const inferenceApi = {
+    enqueue: (sessionId: string, chatId: string, assistantId: string, user: string) => {
+      // Set the inference context so write tools can access sql + sessionId.
+      // The context persists for the duration of the inference run. Since
+      // BooCoder is single-user and runs one inference at a time per session,
+      // this module-level state is safe.
+      setInferenceContext({ sql, sessionId, taskId: null });
+      inference.enqueue(sessionId, chatId, assistantId, user);
+    },
+    cancel: async (sessionId: string, chatId: string) => {
+      const result = await inference.cancel(sessionId, chatId);
+      clearInferenceContext();
+      return result;
+    },
+    hasActive: (chatId: string) => inference.hasActive(chatId),
+  };
+
+  // Register WebSocket support
+  await app.register(fastifyWebsocket);
+
+  // Health endpoint
+  app.get('/api/health', async (_req, reply) => {
+    const dbOk = await pingDb(sql);
+    const status = dbOk ? 200 : 503;
+    return reply.status(status).send({
+      ok: dbOk,
+      db: dbOk,
+      tools: ALL_TOOLS.length,
+    });
+  });
+
+  // Phase 4: probe available agents on startup
+  await probeAgents(sql, app.log);
+
+  // Phase 4: dispatcher — polls tasks table and runs inference
+  const dispatcher = createDispatcher({ sql, inference: inferenceApi, broker, log: app.log, config });
+  dispatcher.start();
+  app.addHook('onClose', () => dispatcher.stop());
+
+  // Register routes
+  registerMessageRoutes(app, sql, broker, inferenceApi);
+  registerPendingRoutes(app, sql);
+  registerTaskRoutes(app, sql, inferenceApi);
+  registerInboxRoutes(app, sql);
+  registerStatsRoutes(app, sql);
+  registerArenaRoutes(app, sql);
+  registerWebSocket(app, sql, broker);
+
+  // Serve static frontend (built web app). In production, the dist/ is
+  // copied to ../web relative to the dist/ directory at /app/web. In dev,
+  // check adjacent to the source.
+  const webRoot = resolve(__dirname, '../web');
+  if (existsSync(webRoot)) {
+    await app.register(fastifyStatic, {
+      root: webRoot,
+      prefix: '/',
+      // Don't intercept /api routes — static only serves files that exist.
+      wildcard: false,
+    });
+    // SPA fallback: serve index.html for non-API routes that don't match a file.
+    app.setNotFoundHandler(async (req, reply) => {
+      if (req.url.startsWith('/api')) {
+        reply.code(404);
+        return { error: 'not found' };
+      }
+      return reply.sendFile('index.html');
+    });
+    app.log.info(`serving frontend from ${webRoot}`);
+  }
+
+  // Graceful shutdown
+  const shutdown = async () => {
+    app.log.info('shutting down');
+    await app.close();
+    await closeDb();
+    process.exit(0);
+  };
+  process.on('SIGTERM', shutdown);
+  process.on('SIGINT', shutdown);
+
+  await app.listen({ port: config.PORT, host: config.HOST });
+  app.log.info(`BooCoder listening on ${config.HOST}:${config.PORT}`);
+}
+
+main().catch((err) => {
+  console.error('fatal:', err);
+  process.exit(1);
+});
--- a/apps/coder/src/routes/arena.ts
+++ b/apps/coder/src/routes/arena.ts
@@ -0,0 +1,122 @@
+/**
+ * v2.0.5: Arena routes — competitive dispatch of the same task to multiple agents.
+ *
+ * POST /api/arena        — create an arena with 2-5 contestants
+ * GET  /api/arena/:id    — get all tasks in an arena
+ * POST /api/arena/:id/select/:task_id — mark a task as the arena winner
+ */
+import type { FastifyInstance } from 'fastify';
+import { z } from 'zod';
+import type { Sql } from '../db.js';
+
+const ContestantSchema = z.object({
+  agent: z.string().max(100).optional(),
+  model: z.string().max(200).optional(),
+});
+
+const CreateArenaBody = z.object({
+  project_id: z.string().uuid(),
+  input: z.string().min(1).max(64_000),
+  contestants: z.array(ContestantSchema).min(2).max(5),
+});
+
+interface TaskRow {
+  id: string;
+  agent: string | null;
+  model: string | null;
+  state: string;
+}
+
+export function registerArenaRoutes(app: FastifyInstance, sql: Sql): void {
+  // POST /api/arena — create a new arena
+  app.post('/api/arena', async (req, reply) => {
+    const parsed = CreateArenaBody.safeParse(req.body);
+    if (!parsed.success) {
+      reply.code(400);
+      return { error: 'invalid body', details: parsed.error.flatten() };
+    }
+
+    const { project_id, input, contestants } = parsed.data;
+    const arenaId = crypto.randomUUID();
+
+    const tasks: TaskRow[] = [];
+    for (const contestant of contestants) {
+      const [task] = await sql<TaskRow[]>`
+        INSERT INTO tasks (project_id, input, agent, model, arena_id)
+        VALUES (${project_id}, ${input}, ${contestant.agent ?? null}, ${contestant.model ?? null}, ${arenaId})
+        RETURNING id, agent, model, state
+      `;
+      tasks.push(task!);
+    }
+
+    reply.code(201);
+    return {
+      arena_id: arenaId,
+      tasks: tasks.map(t => ({
+        id: t.id,
+        agent: t.agent,
+        model: t.model,
+        state: t.state,
+      })),
+    };
+  });
+
+  // GET /api/arena/:arena_id — list all tasks in an arena
+  app.get<{ Params: { arena_id: string } }>('/api/arena/:arena_id', async (req, reply) => {
+    const { arena_id } = req.params;
+
+    // Validate UUID format
+    const uuidRegex = /^[0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12}$/i;
+    if (!uuidRegex.test(arena_id)) {
+      reply.code(400);
+      return { error: 'invalid arena_id format' };
+    }
+
+    const tasks = await sql`
+      SELECT id, project_id, state, input, output_summary, agent, model, execution_path, session_id, started_at, ended_at, created_at, arena_id
+      FROM tasks
+      WHERE arena_id = ${arena_id}
+      ORDER BY created_at
+    `;
+
+    if (tasks.length === 0) {
+      reply.code(404);
+      return { error: 'arena not found' };
+    }
+
+    return { arena_id, tasks };
+  });
+
+  // POST /api/arena/:arena_id/select/:task_id — mark the winner
+  app.post<{ Params: { arena_id: string; task_id: string } }>(
+    '/api/arena/:arena_id/select/:task_id',
+    async (req, reply) => {
+      const { arena_id, task_id } = req.params;
+
+      // Verify the task belongs to this arena
+      const rows = await sql<{ id: string; state: string; arena_id: string | null }[]>`
+        SELECT id, state, arena_id FROM tasks WHERE id = ${task_id}
+      `;
+
+      if (rows.length === 0) {
+        reply.code(404);
+        return { error: 'task not found' };
+      }
+
+      const task = rows[0]!;
+      if (task.arena_id !== arena_id) {
+        reply.code(409);
+        return { error: 'task does not belong to this arena' };
+      }
+
+      // Mark as selected via output_summary prefix (lightweight — no schema change)
+      await sql`
+        UPDATE tasks
+        SET output_summary = COALESCE('[SELECTED] ' || output_summary, '[SELECTED]')
+        WHERE id = ${task_id}
+      `;
+
+      return { selected: true, task_id, arena_id };
+    }
+  );
+}
--- a/apps/coder/src/routes/inbox.ts
+++ b/apps/coder/src/routes/inbox.ts
@@ -0,0 +1,33 @@
+import type { FastifyInstance } from 'fastify';
+import type { Sql } from '../db.js';
+
+export function registerInboxRoutes(app: FastifyInstance, sql: Sql): void {
+  // GET /api/inbox — tasks needing human attention (blocked or failed)
+  app.get('/api/inbox', async () => {
+    return sql`
+      SELECT id, project_id, parent_task_id, state, input, output_summary, agent, model, session_id, started_at, ended_at, created_at
+      FROM human_inbox
+      ORDER BY created_at DESC
+      LIMIT 100
+    `;
+  });
+
+  // POST /api/inbox/:id/retry — reset a blocked/failed task to pending for re-dispatch
+  app.post<{ Params: { id: string } }>('/api/inbox/:id/retry', async (req, reply) => {
+    const taskId = req.params.id;
+
+    const result = await sql`
+      UPDATE tasks
+      SET state = 'pending', started_at = NULL, ended_at = NULL, output_summary = NULL
+      WHERE id = ${taskId} AND state IN ('blocked', 'failed')
+      RETURNING id, state
+    `;
+
+    if (result.length === 0) {
+      reply.code(404);
+      return { error: 'task not found or not in retryable state' };
+    }
+
+    return { id: result[0]!.id, state: result[0]!.state };
+  });
+}
--- a/apps/coder/src/routes/messages.ts
+++ b/apps/coder/src/routes/messages.ts
@@ -0,0 +1,126 @@
+import type { FastifyInstance } from 'fastify';
+import { z } from 'zod';
+import type { Sql } from '../db.js';
+import type { Broker } from '@boocode/server/broker';
+import type { WsFrame } from '@boocode/server/ws-frames';
+
+const SendBody = z.object({
+  content: z.string().min(1).max(64_000),
+  chat_id: z.string().uuid(),
+});
+
+interface InferenceApi {
+  enqueue: (sessionId: string, chatId: string, assistantId: string, user: string) => void;
+  cancel: (sessionId: string, chatId: string) => Promise<boolean>;
+  hasActive: (chatId: string) => boolean;
+}
+
+export function registerMessageRoutes(
+  app: FastifyInstance,
+  sql: Sql,
+  broker: Broker,
+  inference: InferenceApi,
+): void {
+  // POST /api/sessions/:sessionId/messages — send a user message + kick off inference
+  app.post<{ Params: { sessionId: string } }>(
+    '/api/sessions/:sessionId/messages',
+    async (req, reply) => {
+      const parsed = SendBody.safeParse(req.body);
+      if (!parsed.success) {
+        reply.code(400);
+        return { error: 'invalid body', details: parsed.error.flatten() };
+      }
+
+      const sessionId = req.params.sessionId;
+      const { content, chat_id: chatId } = parsed.data;
+
+      // Validate session exists
+      const sessionRows = await sql<{ id: string }[]>`
+        SELECT id FROM sessions WHERE id = ${sessionId}
+      `;
+      if (sessionRows.length === 0) {
+        reply.code(404);
+        return { error: 'session not found' };
+      }
+
+      // Validate chat belongs to session and is open
+      const chatRows = await sql<{ id: string; session_id: string }[]>`
+        SELECT id, session_id FROM chats WHERE id = ${chatId} AND session_id = ${sessionId} AND status = 'open'
+      `;
+      if (chatRows.length === 0) {
+        reply.code(404);
+        return { error: 'chat not found or not open in this session' };
+      }
+
+      // Reject if inference is already running on this chat
+      if (inference.hasActive(chatId)) {
+        reply.code(409);
+        return { error: 'inference already running on this chat' };
+      }
+
+      // Create user message + streaming assistant row in a transaction
+      const result = await sql.begin(async (tx) => {
+        const [userMsg] = await tx<{ id: string }[]>`
+          INSERT INTO messages (session_id, chat_id, role, content, status, created_at)
+          VALUES (${sessionId}, ${chatId}, 'user', ${content}, 'complete', clock_timestamp())
+          RETURNING id
+        `;
+        const [assistantMsg] = await tx<{ id: string }[]>`
+          INSERT INTO messages (session_id, chat_id, role, content, status, created_at)
+          VALUES (${sessionId}, ${chatId}, 'assistant', '', 'streaming', clock_timestamp())
+          RETURNING id
+        `;
+        await tx`UPDATE sessions SET updated_at = clock_timestamp() WHERE id = ${sessionId}`;
+        await tx`UPDATE chats SET updated_at = clock_timestamp() WHERE id = ${chatId}`;
+        return { user_message_id: userMsg!.id, assistant_message_id: assistantMsg!.id };
+      });
+
+      // Publish user message frames so WS subscribers see it immediately
+      broker.publishFrame(sessionId, {
+        type: 'message_started',
+        message_id: result.user_message_id,
+        chat_id: chatId,
+        role: 'user',
+      } as unknown as WsFrame);
+      broker.publishFrame(sessionId, {
+        type: 'delta',
+        message_id: result.user_message_id,
+        chat_id: chatId,
+        content,
+      } as unknown as WsFrame);
+      broker.publishFrame(sessionId, {
+        type: 'message_complete',
+        message_id: result.user_message_id,
+        chat_id: chatId,
+      } as unknown as WsFrame);
+
+      // Enqueue inference — the runner will stream assistant deltas via broker
+      inference.enqueue(sessionId, chatId, result.assistant_message_id, 'default');
+
+      reply.code(202);
+      return result;
+    },
+  );
+
+  // POST /api/sessions/:sessionId/stop — cancel active inference
+  app.post<{ Params: { sessionId: string } }>(
+    '/api/sessions/:sessionId/stop',
+    async (req, reply) => {
+      const sessionId = req.params.sessionId;
+
+      // Find active chats in this session
+      const chats = await sql<{ id: string }[]>`
+        SELECT id FROM chats WHERE session_id = ${sessionId} AND status = 'open'
+      `;
+      let cancelled = false;
+      for (const chat of chats) {
+        if (inference.hasActive(chat.id)) {
+          cancelled = await inference.cancel(sessionId, chat.id);
+          break;
+        }
+      }
+
+      return { cancelled };
+    },
+  );
+}
--- a/apps/coder/src/routes/pending.ts
+++ b/apps/coder/src/routes/pending.ts
@@ -0,0 +1,121 @@
+import type { FastifyInstance } from 'fastify';
+import type { Sql } from '../db.js';
+import {
+  listPending,
+  applyOne,
+  applyAll,
+  rejectOne,
+  rewindOne,
+} from '../services/pending_changes.js';
+
+/**
+ * Resolve project root from a session's project path.
+ */
+async function resolveProjectRoot(sql: Sql, sessionId: string): Promise<string | null> {
+  const rows = await sql<{ path: string }[]>`
+    SELECT p.path FROM sessions s
+    JOIN projects p ON s.project_id = p.id
+    WHERE s.id = ${sessionId}
+  `;
+  return rows.length > 0 ? rows[0]!.path : null;
+}
+
+/**
+ * Resolve project root from a pending change's session.
+ */
+async function resolveProjectRootForChange(sql: Sql, changeId: string): Promise<string | null> {
+  const rows = await sql<{ path: string }[]>`
+    SELECT p.path FROM pending_changes pc
+    JOIN sessions s ON pc.session_id = s.id
+    JOIN projects p ON s.project_id = p.id
+    WHERE pc.id = ${changeId}
+  `;
+  return rows.length > 0 ? rows[0]!.path : null;
+}
+
+export function registerPendingRoutes(app: FastifyInstance, sql: Sql): void {
+  // GET /api/sessions/:sessionId/pending — list pending changes for a session
+  app.get<{ Params: { sessionId: string } }>(
+    '/api/sessions/:sessionId/pending',
+    async (req, reply) => {
+      const sessionId = req.params.sessionId;
+
+      const session = await sql<{ id: string }[]>`SELECT id FROM sessions WHERE id = ${sessionId}`;
+      if (session.length === 0) {
+        reply.code(404);
+        return { error: 'session not found' };
+      }
+
+      const pending = await listPending(sql, sessionId);
+      return pending;
+    },
+  );
+
+  // POST /api/sessions/:sessionId/pending/apply — apply all pending changes
+  app.post<{ Params: { sessionId: string } }>(
+    '/api/sessions/:sessionId/pending/apply',
+    async (req, reply) => {
+      const sessionId = req.params.sessionId;
+
+      const projectRoot = await resolveProjectRoot(sql, sessionId);
+      if (!projectRoot) {
+        reply.code(404);
+        return { error: 'session or project not found' };
+      }
+
+      const results = await applyAll(sql, sessionId, projectRoot);
+      return { results };
+    },
+  );
+
+  // POST /api/pending/:id/apply — apply a single pending change
+  app.post<{ Params: { id: string } }>(
+    '/api/pending/:id/apply',
+    async (req, reply) => {
+      const changeId = req.params.id;
+
+      const projectRoot = await resolveProjectRootForChange(sql, changeId);
+      if (!projectRoot) {
+        reply.code(404);
+        return { error: 'pending change or project not found' };
+      }
+
+      const result = await applyOne(sql, changeId, projectRoot);
+      if (!result.success) {
+        reply.code(422);
+      }
+      return result;
+    },
+  );
+
+  // POST /api/pending/:id/reject — reject a single pending change
+  app.post<{ Params: { id: string } }>(
+    '/api/pending/:id/reject',
+    async (req, reply) => {
+      const changeId = req.params.id;
+
+      await rejectOne(sql, changeId);
+      return { ok: true };
+    },
+  );
+
+  // POST /api/pending/:id/rewind — rewind (undo) an applied change
+  app.post<{ Params: { id: string } }>(
+    '/api/pending/:id/rewind',
+    async (req, reply) => {
+      const changeId = req.params.id;
+
+      const projectRoot = await resolveProjectRootForChange(sql, changeId);
+      if (!projectRoot) {
+        reply.code(404);
+        return { error: 'pending change or project not found' };
+      }
+
+      const result = await rewindOne(sql, changeId, projectRoot);
+      if (!result.success) {
+        reply.code(422);
+      }
+      return result;
+    },
+  );
+}
--- a/apps/coder/src/routes/stats.ts
+++ b/apps/coder/src/routes/stats.ts
@@ -0,0 +1,48 @@
+import type { FastifyInstance } from 'fastify';
+import { z } from 'zod';
+import type { Sql } from '../db.js';
+
+const CostQuery = z.object({
+  group_by: z.enum(['project', 'agent', 'day']).default('project'),
+});
+
+export function registerStatsRoutes(app: FastifyInstance, sql: Sql): void {
+  // GET /api/stats/costs — aggregate cost_tokens by project, agent, or day
+  app.get('/api/stats/costs', async (req, reply) => {
+    const parsed = CostQuery.safeParse(req.query);
+    if (!parsed.success) {
+      reply.code(400);
+      return { error: 'invalid query', details: parsed.error.flatten() };
+    }
+
+    const { group_by } = parsed.data;
+
+    switch (group_by) {
+      case 'project':
+        return sql`
+          SELECT project_id, COUNT(*)::int AS task_count, COALESCE(SUM(cost_tokens), 0)::int AS total_tokens
+          FROM tasks
+          WHERE cost_tokens IS NOT NULL
+          GROUP BY project_id
+          ORDER BY total_tokens DESC
+        `;
+      case 'agent':
+        return sql`
+          SELECT COALESCE(agent, 'native') AS agent, COUNT(*)::int AS task_count, COALESCE(SUM(cost_tokens), 0)::int AS total_tokens
+          FROM tasks
+          WHERE cost_tokens IS NOT NULL
+          GROUP BY agent
+          ORDER BY total_tokens DESC
+        `;
+      case 'day':
+        return sql`
+          SELECT DATE(created_at) AS day, COUNT(*)::int AS task_count, COALESCE(SUM(cost_tokens), 0)::int AS total_tokens
+          FROM tasks
+          WHERE cost_tokens IS NOT NULL
+          GROUP BY DATE(created_at)
+          ORDER BY day DESC
+          LIMIT 90
+        `;
+    }
+  });
+}
--- a/apps/coder/src/routes/tasks.ts
+++ b/apps/coder/src/routes/tasks.ts
@@ -0,0 +1,138 @@
+import type { FastifyInstance } from 'fastify';
+import { z } from 'zod';
+import type { Sql } from '../db.js';
+
+interface InferenceApi {
+  cancel: (sessionId: string, chatId: string) => Promise<boolean>;
+}
+
+const CreateBody = z.object({
+  project_id: z.string().uuid(),
+  input: z.string().min(1).max(64_000),
+  agent: z.string().max(100).optional(),
+  model: z.string().max(200).optional(),
+});
+
+const ListQuery = z.object({
+  state: z.enum(['pending', 'running', 'completed', 'failed', 'blocked', 'cancelled']).optional(),
+  project_id: z.string().uuid().optional(),
+});
+
+export function registerTaskRoutes(app: FastifyInstance, sql: Sql, inference: InferenceApi): void {
+  // POST /api/tasks — create a new task
+  app.post('/api/tasks', async (req, reply) => {
+    const parsed = CreateBody.safeParse(req.body);
+    if (!parsed.success) {
+      reply.code(400);
+      return { error: 'invalid body', details: parsed.error.flatten() };
+    }
+
+    const { project_id, input, agent, model } = parsed.data;
+
+    const [task] = await sql<{ id: string; state: string }[]>`
+      INSERT INTO tasks (project_id, input, agent, model)
+      VALUES (${project_id}, ${input}, ${agent ?? null}, ${model ?? null})
+      RETURNING id, state
+    `;
+
+    reply.code(201);
+    return { id: task!.id, state: task!.state };
+  });
+
+  // GET /api/tasks — list tasks with optional filters
+  app.get('/api/tasks', async (req, _reply) => {
+    const parsed = ListQuery.safeParse(req.query);
+    if (!parsed.success) {
+      return { error: 'invalid query', details: parsed.error.flatten() };
+    }
+
+    const { state, project_id } = parsed.data;
+
+    // Build query with optional filters
+    if (state && project_id) {
+      return sql`
+        SELECT id, project_id, state, input, output_summary, agent, model, execution_path, session_id, started_at, ended_at, created_at
+        FROM tasks
+        WHERE state = ${state} AND project_id = ${project_id}
+        ORDER BY created_at DESC
+        LIMIT 100
+      `;
+    } else if (state) {
+      return sql`
+        SELECT id, project_id, state, input, output_summary, agent, model, execution_path, session_id, started_at, ended_at, created_at
+        FROM tasks
+        WHERE state = ${state}
+        ORDER BY created_at DESC
+        LIMIT 100
+      `;
+    } else if (project_id) {
+      return sql`
+        SELECT id, project_id, state, input, output_summary, agent, model, execution_path, session_id, started_at, ended_at, created_at
+        FROM tasks
+        WHERE project_id = ${project_id}
+        ORDER BY created_at DESC
+        LIMIT 100
+      `;
+    } else {
+      return sql`
+        SELECT id, project_id, state, input, output_summary, agent, model, execution_path, session_id, started_at, ended_at, created_at
+        FROM tasks
+        ORDER BY created_at DESC
+        LIMIT 100
+      `;
+    }
+  });
+
+  // GET /api/tasks/:id — single task detail
+  app.get<{ Params: { id: string } }>('/api/tasks/:id', async (req, reply) => {
+    const rows = await sql`
+      SELECT id, project_id, parent_task_id, state, input, output_summary, agent, model, execution_path, worktree_path, session_id, cost_tokens, started_at, ended_at, created_at
+      FROM tasks
+      WHERE id = ${req.params.id}
+    `;
+    if (rows.length === 0) {
+      reply.code(404);
+      return { error: 'task not found' };
+    }
+    return rows[0];
+  });
+
+  // POST /api/tasks/:id/cancel — cancel a pending or running task
+  app.post<{ Params: { id: string } }>('/api/tasks/:id/cancel', async (req, reply) => {
+    const taskId = req.params.id;
+
+    // Get current task state + session info
+    const rows = await sql<{ id: string; state: string; session_id: string | null }[]>`
+      SELECT id, state, session_id FROM tasks WHERE id = ${taskId}
+    `;
+    if (rows.length === 0) {
+      reply.code(404);
+      return { error: 'task not found' };
+    }
+
+    const task = rows[0]!;
+    if (task.state !== 'pending' && task.state !== 'running') {
+      reply.code(409);
+      return { error: `cannot cancel task in state '${task.state}'` };
+    }
+
+    // If running, try to cancel inference
+    if (task.state === 'running' && task.session_id) {
+      // Find active chat in the task's session
+      const chats = await sql<{ id: string }[]>`
+        SELECT id FROM chats WHERE session_id = ${task.session_id} AND status = 'open'
+      `;
+      for (const chat of chats) {
+        await inference.cancel(task.session_id, chat.id);
+      }
+    }
+
+    await sql`
+      UPDATE tasks
+      SET state = 'cancelled', ended_at = clock_timestamp()
+      WHERE id = ${taskId} AND state IN ('pending', 'running')
+    `;
+
+    return { cancelled: true };
+  });
+}
--- a/apps/coder/src/routes/ws.ts
+++ b/apps/coder/src/routes/ws.ts
@@ -0,0 +1,51 @@
+import type { FastifyInstance } from 'fastify';
+import type { Sql } from '../db.js';
+import type { Broker } from '@boocode/server/broker';
+
+export function registerWebSocket(
+  app: FastifyInstance,
+  sql: Sql,
+  broker: Broker,
+): void {
+  // Per-session streaming WebSocket. Clients connect here to receive live
+  // inference frames (deltas, tool_calls, tool_results, message_complete).
+  app.get<{ Params: { sessionId: string } }>(
+    '/api/ws/sessions/:sessionId',
+    { websocket: true },
+    async (socket, req) => {
+      const sessionId = req.params.sessionId;
+
+      // Validate session exists
+      const session = await sql<{ id: string }[]>`SELECT id FROM sessions WHERE id = ${sessionId}`;
+      if (session.length === 0) {
+        socket.send(JSON.stringify({ type: 'error', error: 'session not found' }));
+        socket.close(1008, 'session not found');
+        return;
+      }
+
+      // Send snapshot of existing messages so client can hydrate
+      const messages = await sql<Record<string, unknown>[]>`
+        SELECT id, session_id, chat_id, role, content, kind, tool_calls, tool_results, status, last_seq,
+               tokens_used, ctx_used, ctx_max, started_at, finished_at, created_at, metadata,
+               summary, tail_start_id, compacted_at
+        FROM messages_with_parts
+        WHERE session_id = ${sessionId}
+        ORDER BY created_at ASC, id ASC
+      `;
+      socket.send(JSON.stringify({ type: 'snapshot', messages }));
+
+      // Subscribe to broker for live frames
+      const unsubscribe = broker.subscribe(sessionId, (frame) => {
+        if (socket.readyState !== socket.OPEN) return;
+        try {
+          socket.send(JSON.stringify(frame));
+        } catch (err) {
+          app.log.warn({ err, sessionId }, 'ws send failed');
+        }
+      });
+
+      socket.on('close', () => unsubscribe());
+      socket.on('error', () => unsubscribe());
+    },
+  );
+}
--- a/apps/coder/src/schema.sql
+++ b/apps/coder/src/schema.sql
@@ -0,0 +1,63 @@
+-- v2.0.0: BooCoder schema — pending changes, tasks, agent registry.
+-- Applied on startup by apps/coder/src/db.ts:applySchema().
+-- Lives in the same 'boochat' database as BooChat's tables.
+
+CREATE TABLE IF NOT EXISTS pending_changes (
+  id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
+  session_id UUID NOT NULL,
+  task_id UUID,
+  file_path TEXT NOT NULL,
+  operation TEXT NOT NULL,
+  diff TEXT NOT NULL,
+  status TEXT NOT NULL DEFAULT 'pending',
+  created_at TIMESTAMPTZ NOT NULL DEFAULT clock_timestamp(),
+  CONSTRAINT pending_changes_operation_chk CHECK (operation IN ('create', 'edit', 'delete')),
+  CONSTRAINT pending_changes_status_chk CHECK (status IN ('pending', 'applied', 'rejected', 'reverted'))
+);
+
+CREATE TABLE IF NOT EXISTS tasks (
+  id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
+  project_id UUID NOT NULL,
+  parent_task_id UUID REFERENCES tasks(id),
+  state TEXT NOT NULL DEFAULT 'pending',
+  input TEXT NOT NULL,
+  output_summary TEXT,
+  agent TEXT,
+  model TEXT,
+  execution_path TEXT,
+  worktree_path TEXT,
+  cost_tokens INTEGER,
+  started_at TIMESTAMPTZ,
+  ended_at TIMESTAMPTZ,
+  created_at TIMESTAMPTZ NOT NULL DEFAULT clock_timestamp(),
+  CONSTRAINT tasks_state_chk CHECK (state IN ('pending', 'running', 'completed', 'failed', 'blocked', 'cancelled')),
+  CONSTRAINT tasks_execution_path_chk CHECK (execution_path IS NULL OR execution_path IN ('native', 'acp', 'pty', 'qwen'))
+);
+
+CREATE TABLE IF NOT EXISTS available_agents (
+  name TEXT PRIMARY KEY,
+  install_path TEXT,
+  version TEXT,
+  supports_acp BOOLEAN NOT NULL DEFAULT false,
+  supports_mcp_client BOOLEAN NOT NULL DEFAULT false,
+  last_probed_at TIMESTAMPTZ
+);
+
+-- v2.0.0 Phase 4: link tasks to their inference sessions.
+ALTER TABLE tasks ADD COLUMN IF NOT EXISTS session_id UUID REFERENCES sessions(id);
+
+-- v2.0.5: add 'qwen' to execution_path CHECK + arena_id column.
+ALTER TABLE tasks DROP CONSTRAINT IF EXISTS tasks_execution_path_chk;
+DO $$ BEGIN
+  IF NOT EXISTS (SELECT 1 FROM pg_constraint WHERE conname = 'tasks_execution_path_chk') THEN
+    ALTER TABLE tasks ADD CONSTRAINT tasks_execution_path_chk
+      CHECK (execution_path IS NULL OR execution_path IN ('native', 'acp', 'pty', 'qwen'));
+  END IF;
+END $$;
+
+-- v2.0.5: arena support — group tasks into competitive arenas.
+ALTER TABLE tasks ADD COLUMN IF NOT EXISTS arena_id UUID;
+
+-- Human inbox: tasks needing attention
+CREATE OR REPLACE VIEW human_inbox AS
+  SELECT * FROM tasks WHERE state IN ('blocked', 'failed');
--- a/apps/coder/src/services/tests/pending_changes_integration.test.ts
+++ b/apps/coder/src/services/tests/pending_changes_integration.test.ts
@@ -0,0 +1,96 @@
+import { describe, it, expect, beforeAll, afterAll } from 'vitest';
+import { readFileSync, existsSync } from 'node:fs';
+import { readFile, rm, mkdir } from 'node:fs/promises';
+import { resolve } from 'node:path';
+import postgres from 'postgres';
+import { queueCreate, queueEdit, queueDelete, applyOne, rewindOne, listPending } from '../pending_changes.js';
+
+/**
+ * Integration test for the full pending-changes lifecycle.
+ * Requires DATABASE_URL env var pointing to a running postgres instance.
+ * Skips cleanly when DATABASE_URL is not set.
+ *
+ * Run with:
+ *   DATABASE_URL='postgres://boocode:devpass@localhost:5500/boocode' pnpm -C apps/coder test
+ */
+describe.runIf(!!process.env.DATABASE_URL)('pending_changes integration', () => {
+  let sql: ReturnType<typeof postgres>;
+  const testDir = '/tmp/boocode-pending-changes-test-' + Date.now();
+  const projectRoot = testDir;
+  const testSessionId = '00000000-0000-0000-0000-000000000001';
+
+  beforeAll(async () => {
+    sql = postgres(process.env.DATABASE_URL!, { max: 3 });
+
+    // Apply schema
+    const schemaPath = resolve(__dirname, '../../schema.sql');
+    const ddl = readFileSync(schemaPath, 'utf8');
+    await sql.unsafe(ddl);
+
+    // Create temp project directory
+    await mkdir(testDir, { recursive: true });
+  });
+
+  afterAll(async () => {
+    // Cleanup test data
+    await sql`DELETE FROM pending_changes WHERE session_id = ${testSessionId}`;
+    await sql.end({ timeout: 5 });
+    // Remove temp directory
+    await rm(testDir, { recursive: true, force: true });
+  });
+
+  it('queueCreate → listPending → applyOne → verify file exists', async () => {
+    const change = await queueCreate(sql, testSessionId, null, 'hello.txt', 'hello world', projectRoot);
+    expect(change.status).toBe('pending');
+    expect(change.operation).toBe('create');
+
+    const pending = await listPending(sql, testSessionId);
+    expect(pending.some((p) => p.id === change.id)).toBe(true);
+
+    const result = await applyOne(sql, change.id, projectRoot);
+    expect(result.success).toBe(true);
+
+    const content = await readFile(resolve(testDir, 'hello.txt'), 'utf8');
+    expect(content).toBe('hello world');
+  });
+
+  it('queueEdit → apply → verify content changed', async () => {
+    // Setup: create a file first
+    const createChange = await queueCreate(sql, testSessionId, null, 'editable.txt', 'original content here', projectRoot);
+    await applyOne(sql, createChange.id, projectRoot);
+
+    // Queue an edit
+    const editChange = await queueEdit(sql, testSessionId, null, 'editable.txt', 'original', 'modified', projectRoot);
+    expect(editChange.operation).toBe('edit');
+
+    const result = await applyOne(sql, editChange.id, projectRoot);
+    expect(result.success).toBe(true);
+
+    const content = await readFile(resolve(testDir, 'editable.txt'), 'utf8');
+    expect(content).toBe('modified content here');
+  });
+
+  it('queueDelete → apply → verify file gone', async () => {
+    // Setup: create a file
+    const createChange = await queueCreate(sql, testSessionId, null, 'deleteme.txt', 'goodbye', projectRoot);
+    await applyOne(sql, createChange.id, projectRoot);
+    expect(existsSync(resolve(testDir, 'deleteme.txt'))).toBe(true);
+
+    // Queue a delete
+    const deleteChange = await queueDelete(sql, testSessionId, null, 'deleteme.txt', projectRoot);
+    const result = await applyOne(sql, deleteChange.id, projectRoot);
+    expect(result.success).toBe(true);
+    expect(existsSync(resolve(testDir, 'deleteme.txt'))).toBe(false);
+  });
+
+  it('rewindOne → verify reverted', async () => {
+    // Setup: create and apply a file
+    const createChange = await queueCreate(sql, testSessionId, null, 'rewindable.txt', 'initial', projectRoot);
+    await applyOne(sql, createChange.id, projectRoot);
+
+    // Rewind the create (should delete the file)
+    const result = await rewindOne(sql, createChange.id, projectRoot);
+    expect(result.success).toBe(true);
+    expect(existsSync(resolve(testDir, 'rewindable.txt'))).toBe(false);
+  });
+});
--- a/apps/coder/src/services/tests/write_guard.test.ts
+++ b/apps/coder/src/services/tests/write_guard.test.ts
@@ -0,0 +1,115 @@
+import { describe, it, expect } from 'vitest';
+import { resolveWritePath, isSecretPath, WriteGuardError } from '../write_guard.js';
+
+const PROJECT_ROOT = '/opt/projects/my-app';
+
+describe('resolveWritePath', () => {
+  it('resolves a relative path correctly', () => {
+    const result = resolveWritePath(PROJECT_ROOT, 'src/index.ts');
+    expect(result).toBe('/opt/projects/my-app/src/index.ts');
+  });
+
+  it('resolves nested relative path', () => {
+    const result = resolveWritePath(PROJECT_ROOT, 'src/lib/utils.ts');
+    expect(result).toBe('/opt/projects/my-app/src/lib/utils.ts');
+  });
+
+  it('throws on ../ escape', () => {
+    expect(() => resolveWritePath(PROJECT_ROOT, '../../../etc/passwd')).toThrow(WriteGuardError);
+    expect(() => resolveWritePath(PROJECT_ROOT, '../../../etc/passwd')).toThrow('path escapes project root');
+  });
+
+  it('throws on absolute path outside project root', () => {
+    expect(() => resolveWritePath(PROJECT_ROOT, '/etc/shadow')).toThrow(WriteGuardError);
+    expect(() => resolveWritePath(PROJECT_ROOT, '/tmp/exploit')).toThrow('path escapes project root');
+  });
+
+  it('allows absolute path inside project root', () => {
+    const result = resolveWritePath(PROJECT_ROOT, '/opt/projects/my-app/src/new.ts');
+    expect(result).toBe('/opt/projects/my-app/src/new.ts');
+  });
+
+  it('denies .env files', () => {
+    expect(() => resolveWritePath(PROJECT_ROOT, '.env')).toThrow(WriteGuardError);
+    expect(() => resolveWritePath(PROJECT_ROOT, '.env')).toThrow('cannot write to secret file');
+  });
+
+  it('denies .env.local', () => {
+    expect(() => resolveWritePath(PROJECT_ROOT, '.env.local')).toThrow(WriteGuardError);
+  });
+
+  it('denies .env.production', () => {
+    expect(() => resolveWritePath(PROJECT_ROOT, '.env.production')).toThrow(WriteGuardError);
+  });
+
+  it('denies *.pem files', () => {
+    expect(() => resolveWritePath(PROJECT_ROOT, 'certs/server.pem')).toThrow(WriteGuardError);
+    expect(() => resolveWritePath(PROJECT_ROOT, 'certs/server.pem')).toThrow('cannot write to secret file');
+  });
+
+  it('denies *.key files', () => {
+    expect(() => resolveWritePath(PROJECT_ROOT, 'ssl/private.key')).toThrow(WriteGuardError);
+  });
+
+  it('denies id_rsa', () => {
+    expect(() => resolveWritePath(PROJECT_ROOT, '.ssh/id_rsa')).toThrow(WriteGuardError);
+  });
+
+  it('denies id_ed25519', () => {
+    expect(() => resolveWritePath(PROJECT_ROOT, '.ssh/id_ed25519')).toThrow(WriteGuardError);
+  });
+
+  it('denies credentials.json', () => {
+    expect(() => resolveWritePath(PROJECT_ROOT, 'credentials.json')).toThrow(WriteGuardError);
+  });
+
+  it('passes a normal file inside project', () => {
+    const result = resolveWritePath(PROJECT_ROOT, 'src/components/Button.tsx');
+    expect(result).toBe('/opt/projects/my-app/src/components/Button.tsx');
+  });
+
+  it('passes a non-existent nested file (no realpath)', () => {
+    // This is the key difference from BooChat's pathGuard: no realpath means
+    // files that don't exist yet still pass validation
+    const result = resolveWritePath(PROJECT_ROOT, 'src/new-dir/new-file.ts');
+    expect(result).toBe('/opt/projects/my-app/src/new-dir/new-file.ts');
+  });
+
+  it('throws on null/empty path', () => {
+    expect(() => resolveWritePath(PROJECT_ROOT, '')).toThrow(WriteGuardError);
+    expect(() => resolveWritePath(PROJECT_ROOT, '')).toThrow('file path is required');
+  });
+
+  it('normalizes ../ within project root and still allows', () => {
+    const result = resolveWritePath(PROJECT_ROOT, 'src/../lib/utils.ts');
+    expect(result).toBe('/opt/projects/my-app/lib/utils.ts');
+  });
+
+  it('rejects path that looks inside root but normalizes outside', () => {
+    expect(() => resolveWritePath(PROJECT_ROOT, 'src/../../other-project/hack.ts')).toThrow(WriteGuardError);
+  });
+});
+
+describe('isSecretPath', () => {
+  it('detects .env', () => {
+    expect(isSecretPath('.env')).toBe(true);
+  });
+
+  it('detects nested .env', () => {
+    expect(isSecretPath('config/.env')).toBe(true);
+  });
+
+  it('detects *.pfx', () => {
+    expect(isSecretPath('certs/client.pfx')).toBe(true);
+  });
+
+  it('does not flag normal source files', () => {
+    expect(isSecretPath('src/index.ts')).toBe(false);
+    expect(isSecretPath('README.md')).toBe(false);
+    expect(isSecretPath('package.json')).toBe(false);
+  });
+
+  it('returns false for empty string', () => {
+    expect(isSecretPath('')).toBe(false);
+  });
+});
--- a/apps/coder/src/services/tests/write_guard_fuzz.test.ts
+++ b/apps/coder/src/services/tests/write_guard_fuzz.test.ts
@@ -0,0 +1,193 @@
+import { describe, it, expect } from 'vitest';
+import { resolveWritePath } from '../write_guard.js';
+
+const projectRoot = '/opt/testproject';
+
+describe('write_guard fuzz — traversal attacks', () => {
+  // Basic traversal
+  it('rejects ../', () => {
+    expect(() => resolveWritePath(projectRoot, '../etc/passwd')).toThrow();
+  });
+
+  it('rejects ../../', () => {
+    expect(() => resolveWritePath(projectRoot, '../../etc/passwd')).toThrow();
+  });
+
+  it('rejects deeply nested ../../../', () => {
+    expect(() => resolveWritePath(projectRoot, '../../../../../../../etc/shadow')).toThrow();
+  });
+
+  // Encoded traversal — resolve() doesn't decode percent-encoding, so these
+  // stay as literal filenames. The guard must still not let them escape.
+  it('rejects %2e%2e/ (literal percent-encoded dots)', () => {
+    // resolve('/opt/testproject', '%2e%2e/etc/passwd') stays inside root
+    // because Node's resolve treats the literal characters, not decoded.
+    // The file would be /opt/testproject/%2e%2e/etc/passwd which IS inside root.
+    // This test confirms it doesn't throw (it resolves inside) — defense in depth
+    // is that the filesystem won't have this path, but no traversal occurs.
+    const result = resolveWritePath(projectRoot, '%2e%2e/etc/passwd');
+    expect(result).toContain(projectRoot);
+  });
+
+  it('rejects ..%2f (literal percent-encoded slash)', () => {
+    // '../%2fetc/passwd' — the ../ IS real traversal
+    expect(() => resolveWritePath(projectRoot, '../%2fetc/passwd')).toThrow();
+  });
+
+  // Null byte injection
+  it('rejects null bytes', () => {
+    expect(() => resolveWritePath(projectRoot, 'file.txt\x00.jpg')).toThrow();
+  });
+
+  // Absolute path escape
+  it('rejects /etc/passwd', () => {
+    expect(() => resolveWritePath(projectRoot, '/etc/passwd')).toThrow();
+  });
+
+  it('rejects /opt/other-project/file', () => {
+    expect(() => resolveWritePath(projectRoot, '/opt/other-project/file.ts')).toThrow();
+  });
+
+  // Path that starts with project root as prefix but isn't under it
+  it('rejects prefix match without separator', () => {
+    expect(() => resolveWritePath(projectRoot, '/opt/testproject-evil/file.ts')).toThrow();
+  });
+
+  // Double slashes / traversal after valid prefix
+  it('rejects /opt/testproject/../etc/passwd via double-dot after valid prefix', () => {
+    expect(() => resolveWritePath(projectRoot, '/opt/testproject/../etc/passwd')).toThrow();
+  });
+
+  // Windows-style (defense-in-depth on Linux)
+  it('rejects backslash traversal', () => {
+    // On POSIX, backslash is a valid filename char, so '..\\etc\\passwd' resolves
+    // as a single segment inside projectRoot. Not a traversal, but test that it
+    // doesn't crash and stays within root.
+    const result = resolveWritePath(projectRoot, '..\\etc\\passwd');
+    // Node resolve on POSIX treats this as a literal filename segment containing backslashes
+    // that starts with '..' — resolve normalizes: /opt/testproject/..\\etc\\passwd
+    // Wait: resolve('/opt/testproject', '..\\etc\\passwd') — on POSIX backslash
+    // is NOT a separator, so this is a file named '..\\etc\\passwd' inside projectRoot.
+    // Actually no — resolve splits on '/' only on POSIX. '..' at start triggers parent.
+    // Let's check: the string starts with '..' but the next char is '\\' not '/'.
+    // Node's path.resolve on POSIX: the string '..\\etc\\passwd' does NOT contain '/'
+    // so it IS treated as a single path component? No — resolve still splits on '/'.
+    // '..\\etc\\passwd' has no '/', so resolve('/opt/testproject', '..\\etc\\passwd')
+    // = resolve('/opt/testproject/..\\etc\\passwd') — but wait, resolve processes
+    // segments separated by '/'. With no '/', the whole thing is one segment.
+    // Actually wrong: path.resolve calls normalizeString which handles '.' and '..'
+    // only when they are full segments delimited by '/'. Since there's no '/' in
+    // '..\\etc\\passwd', it treats the entire string as one filename.
+    // So: /opt/testproject/..\\etc\\passwd — inside root. No throw.
+    expect(result).toContain(projectRoot);
+  });
+
+  // Secret files (deny list)
+  it('rejects .env', () => {
+    expect(() => resolveWritePath(projectRoot, '.env')).toThrow();
+  });
+
+  it('rejects nested .env', () => {
+    expect(() => resolveWritePath(projectRoot, 'config/.env')).toThrow();
+  });
+
+  it('rejects .env.local', () => {
+    expect(() => resolveWritePath(projectRoot, '.env.local')).toThrow();
+  });
+
+  it('rejects id_rsa', () => {
+    expect(() => resolveWritePath(projectRoot, '.ssh/id_rsa')).toThrow();
+  });
+
+  it('rejects id_ed25519', () => {
+    expect(() => resolveWritePath(projectRoot, '.ssh/id_ed25519')).toThrow();
+  });
+
+  it('rejects *.pem', () => {
+    expect(() => resolveWritePath(projectRoot, 'certs/server.pem')).toThrow();
+  });
+
+  it('rejects *.key', () => {
+    expect(() => resolveWritePath(projectRoot, 'certs/private.key')).toThrow();
+  });
+
+  it('rejects credentials.json', () => {
+    expect(() => resolveWritePath(projectRoot, 'credentials.json')).toThrow();
+  });
+
+  it('rejects *.p12', () => {
+    expect(() => resolveWritePath(projectRoot, 'certs/client.p12')).toThrow();
+  });
+
+  it('rejects .netrc', () => {
+    expect(() => resolveWritePath(projectRoot, '.netrc')).toThrow();
+  });
+
+  it('rejects *.kdbx', () => {
+    expect(() => resolveWritePath(projectRoot, 'secrets/passwords.kdbx')).toThrow();
+  });
+
+  // Valid paths (should NOT throw)
+  it('allows simple relative path', () => {
+    expect(resolveWritePath(projectRoot, 'src/index.ts')).toBe('/opt/testproject/src/index.ts');
+  });
+
+  it('allows nested path', () => {
+    expect(resolveWritePath(projectRoot, 'src/services/tools/edit_file.ts')).toContain(projectRoot);
+  });
+
+  it('allows dotfile that is not in deny list', () => {
+    expect(resolveWritePath(projectRoot, '.gitignore')).toContain(projectRoot);
+  });
+
+  it('allows absolute path inside project', () => {
+    expect(resolveWritePath(projectRoot, '/opt/testproject/new-file.ts')).toBe('/opt/testproject/new-file.ts');
+  });
+
+  it('allows path with safe internal ../', () => {
+    expect(resolveWritePath(projectRoot, 'src/../lib/utils.ts')).toBe('/opt/testproject/lib/utils.ts');
+  });
+});
+
+describe('write_guard fuzz — edge cases', () => {
+  it('throws on empty string', () => {
+    expect(() => resolveWritePath(projectRoot, '')).toThrow();
+  });
+
+  it('throws on whitespace-only', () => {
+    expect(() => resolveWritePath(projectRoot, '   ')).toThrow();
+  });
+
+  it('throws when path IS the project root itself', () => {
+    // Writing to the directory itself makes no sense for a file write
+    expect(() => resolveWritePath(projectRoot, '/opt/testproject')).not.toThrow();
+    // The guard allows it (resolve === projectRoot passes the check).
+    // This is acceptable because the filesystem write will fail on a directory.
+    // If we want to block this, that's a separate concern.
+  });
+
+  it('handles very long path without crashing', () => {
+    const longSegment = 'a'.repeat(255);
+    const longPath = Array(20).fill(longSegment).join('/');
+    // Should not crash — may throw or succeed, but must not buffer-overflow
+    expect(() => resolveWritePath(projectRoot, longPath)).not.toThrow();
+  });
+
+  it('handles path with only dots', () => {
+    // Single dot resolves to projectRoot itself
+    const result = resolveWritePath(projectRoot, './src/file.ts');
+    expect(result).toBe('/opt/testproject/src/file.ts');
+  });
+
+  it('rejects triple-dot trick (... is not special but ../ within is)', () => {
+    // '.../etc' is a literal directory name, not traversal
+    const result = resolveWritePath(projectRoot, '.../etc');
+    expect(result).toContain(projectRoot);
+  });
+
+  it('rejects path with multiple consecutive slashes', () => {
+    // resolve normalizes these; should still be inside root
+    const result = resolveWritePath(projectRoot, 'src///file.ts');
+    expect(result).toBe('/opt/testproject/src/file.ts');
+  });
+});
--- a/apps/coder/src/services/acp-dispatch.ts
+++ b/apps/coder/src/services/acp-dispatch.ts
@@ -0,0 +1,271 @@
+/**
+ * ACP dispatch — runs ACP-capable agents (opencode, goose) on the host via SSH.
+ *
+ * Uses the @agentclientprotocol/sdk to establish a structured JSON-RPC session
+ * with the agent subprocess. The SSH tunnel provides stdio transport.
+ *
+ * Flow:
+ * 1. SSH to host, start `opencode acp` (or `goose acp`) in the worktree
+ * 2. Wrap SSH child's stdin/stdout into NDJSON streams
+ * 3. Create a ClientSideConnection from the SDK
+ * 4. Initialize → newSession → prompt(task)
+ * 5. Collect session updates (tool calls, text output)
+ * 6. On prompt completion → return collected output
+ */
+import type { FastifyBaseLogger } from 'fastify';
+import { Readable, Writable } from 'node:stream';
+import {
+  ClientSideConnection,
+  ndJsonStream,
+  type Client,
+  type SessionNotification,
+  type RequestPermissionRequest,
+  type RequestPermissionResponse,
+  type ReadTextFileRequest,
+  type ReadTextFileResponse,
+  type WriteTextFileRequest,
+  type WriteTextFileResponse,
+  type CreateTerminalRequest,
+  type CreateTerminalResponse,
+} from '@agentclientprotocol/sdk';
+import { sshSpawn } from './ssh.js';
+
+export interface AcpDispatchResult {
+  exitCode: number;
+  output: string;
+  toolCalls: Array<{ title: string; input: unknown; output?: unknown }>;
+  stopReason: string;
+}
+
+export interface AcpDispatchOpts {
+  agent: string;
+  task: string;
+  worktreePath: string;
+  model?: string;
+  signal?: AbortSignal;
+  log: FastifyBaseLogger;
+}
+
+/** Map agent name to the ACP command it exposes. */
+function acpCommand(agent: string): string | null {
+  switch (agent) {
+    case 'opencode':
+      return 'opencode acp';
+    case 'goose':
+      return 'goose acp';
+    default:
+      return null;
+  }
+}
+
+/**
+ * Convert a Node.js Readable stream to a web ReadableStream<Uint8Array>.
+ */
+function nodeReadableToWeb(nodeStream: NodeJS.ReadableStream): ReadableStream<Uint8Array> {
+  return new ReadableStream<Uint8Array>({
+    start(controller) {
+      nodeStream.on('data', (chunk: Buffer) => {
+        controller.enqueue(new Uint8Array(chunk));
+      });
+      nodeStream.on('end', () => {
+        controller.close();
+      });
+      nodeStream.on('error', (err) => {
+        controller.error(err);
+      });
+    },
+    cancel() {
+      if ('destroy' in nodeStream && typeof (nodeStream as Readable).destroy === 'function') {
+        (nodeStream as Readable).destroy();
+      }
+    },
+  });
+}
+
+/**
+ * Convert a Node.js Writable stream to a web WritableStream<Uint8Array>.
+ */
+function nodeWritableToWeb(nodeStream: NodeJS.WritableStream): WritableStream<Uint8Array> {
+  return new WritableStream<Uint8Array>({
+    write(chunk) {
+      return new Promise<void>((resolve, reject) => {
+        const ok = (nodeStream as Writable).write(chunk, (err) => {
+          if (err) reject(err);
+        });
+        if (ok) resolve();
+        else (nodeStream as Writable).once('drain', resolve);
+      });
+    },
+    close() {
+      return new Promise<void>((resolve) => {
+        (nodeStream as Writable).end(resolve);
+      });
+    },
+    abort() {
+      (nodeStream as Writable).destroy();
+    },
+  });
+}
+
+/**
+ * Dispatch a task to an ACP-capable agent via SSH.
+ *
+ * Opens a structured ACP session, sends the task as a prompt, and collects
+ * all session updates. Returns the collected output and tool calls.
+ */
+export async function dispatchViaAcp(opts: AcpDispatchOpts): Promise<AcpDispatchResult> {
+  const { agent, task, worktreePath, signal, log } = opts;
+
+  const cmd = acpCommand(agent);
+  if (!cmd) {
+    return {
+      exitCode: 1,
+      output: `Agent '${agent}' does not support ACP.`,
+      toolCalls: [],
+      stopReason: 'error',
+    };
+  }
+
+  // Spawn SSH with the ACP command running in the worktree
+  const escapedPath = worktreePath.replace(/'/g, "'\\''");
+  const fullCommand = `cd '${escapedPath}' && ${cmd}`;
+
+  log.info({ agent, worktreePath }, 'acp-dispatch: spawning');
+  const child = sshSpawn(fullCommand);
+
+  // Wire up abort
+  let killed = false;
+  const cleanup = () => {
+    if (!killed) {
+      killed = true;
+      child.kill('SIGTERM');
+      setTimeout(() => child.kill('SIGKILL'), 5_000);
+    }
+  };
+
+  if (signal) {
+    if (signal.aborted) {
+      cleanup();
+      return { exitCode: 130, output: 'Aborted before start', toolCalls: [], stopReason: 'cancelled' };
+    }
+    signal.addEventListener('abort', cleanup, { once: true });
+  }
+
+  try {
+    // Create web streams from the child process stdio
+    const inputStream = nodeReadableToWeb(child.stdout!);
+    const outputStream = nodeWritableToWeb(child.stdin!);
+
+    // Create the NDJSON ACP stream
+    const stream = ndJsonStream(outputStream, inputStream);
+
+    // Collected session updates
+    const textChunks: string[] = [];
+    const toolCalls: Array<{ title: string; input: unknown; output?: unknown }> = [];
+
+    // Create client-side connection — we are the "client" (editor), the agent is remote
+    const connection = new ClientSideConnection(
+      (_agentInterface): Client => ({
+        // Handle session updates from the agent
+        async sessionUpdate(params: SessionNotification): Promise<void> {
+          const update = params.update;
+          if (update.sessionUpdate === 'agent_message_chunk') {
+            // ContentChunk with content: ContentBlock
+            const content = update.content;
+            if (content.type === 'text' && 'text' in content) {
+              textChunks.push((content as { text: string }).text);
+            }
+          } else if (update.sessionUpdate === 'tool_call') {
+            toolCalls.push({
+              title: update.title,
+              input: update.rawInput,
+            });
+          } else if (update.sessionUpdate === 'tool_call_update') {
+            const last = toolCalls[toolCalls.length - 1];
+            if (last && update.rawOutput !== undefined) {
+              last.output = update.rawOutput;
+            }
+          }
+        },
+
+        // Permission requests — auto-approve by selecting the first option (worktree is isolated)
+        async requestPermission(params: RequestPermissionRequest): Promise<RequestPermissionResponse> {
+          // Select the first available option to auto-approve
+          const firstOption = params.options[0];
+          if (firstOption) {
+            return {
+              outcome: { outcome: 'selected', optionId: firstOption.optionId },
+            };
+          }
+          // No options available — cancel
+          return { outcome: { outcome: 'cancelled' } };
+        },
+
+        // File system operations — let the agent handle them directly in the worktree
+        async readTextFile(_params: ReadTextFileRequest): Promise<ReadTextFileResponse> {
+          return { content: '' };
+        },
+        async writeTextFile(_params: WriteTextFileRequest): Promise<WriteTextFileResponse> {
+          return {};
+        },
+        async createTerminal(_params: CreateTerminalRequest): Promise<CreateTerminalResponse> {
+          return { terminalId: 'noop' };
+        },
+      }),
+      stream,
+    );
+
+    // Initialize the connection
+    // ProtocolVersion is a number in this SDK version
+    const initResult = await connection.initialize({
+      protocolVersion: 1,
+      clientInfo: { name: 'boocoder', version: '2.0.1' },
+      clientCapabilities: {},
+    });
+    log.info({ agentInfo: initResult.agentInfo }, 'acp-dispatch: initialized');
+
+    // Create a new session
+    const session = await connection.newSession({
+      cwd: worktreePath,
+      mcpServers: [],
+    });
+    log.info({ sessionId: session.sessionId }, 'acp-dispatch: session created');
+
+    // Send the prompt
+    const promptResult = await connection.prompt({
+      sessionId: session.sessionId,
+      prompt: [{ type: 'text', text: task }],
+    });
+
+    const stopReason = promptResult.stopReason ?? 'end_turn';
+    log.info({ agent, stopReason, toolCallCount: toolCalls.length }, 'acp-dispatch: prompt completed');
+
+    // Clean shutdown
+    await connection.closeSession({ sessionId: session.sessionId }).catch(() => {});
+
+    return {
+      exitCode: 0,
+      output: textChunks.join(''),
+      toolCalls,
+      stopReason,
+    };
+  } catch (err) {
+    const message = err instanceof Error ? err.message : String(err);
+    log.error({ agent, err: message }, 'acp-dispatch: error');
+    return {
+      exitCode: 1,
+      output: message,
+      toolCalls: [],
+      stopReason: 'error',
+    };
+  } finally {
+    if (signal) signal.removeEventListener('abort', cleanup);
+    cleanup();
+
+    // Wait for child to exit
+    await new Promise<void>((resolve) => {
+      child.on('close', resolve);
+      setTimeout(resolve, 3_000);
+    });
+  }
+}
--- a/apps/coder/src/services/agent-probe.ts
+++ b/apps/coder/src/services/agent-probe.ts
@@ -0,0 +1,71 @@
+import type { Sql } from '../db.js';
+import type { FastifyBaseLogger } from 'fastify';
+import { sshExec } from './ssh.js';
+
+const KNOWN_AGENTS: Array<{ name: string; supportsAcp: boolean }> = [
+  { name: 'opencode', supportsAcp: true },
+  { name: 'goose', supportsAcp: true },
+  { name: 'claude', supportsAcp: false },
+  { name: 'pi', supportsAcp: false },
+  { name: 'qwen', supportsAcp: false },
+];
+
+/**
+ * Probe for available agents on the HOST via SSH.
+ *
+ * The boocoder container can't run agents locally — they live on the host.
+ * We SSH to the host (same mechanism BooTerm uses) and check which agent
+ * binaries are on PATH.
+ */
+export async function probeAgents(sql: Sql, log: FastifyBaseLogger): Promise<void> {
+  log.info('agent-probe: scanning HOST for known agents via SSH');
+
+  for (const agent of KNOWN_AGENTS) {
+    try {
+      // Check if the agent binary is on the host's PATH
+      const whichResult = await sshExec(`which ${agent.name}`, { timeoutMs: 10_000 });
+      const installPath = whichResult.stdout.trim();
+      if (whichResult.exitCode !== 0 || !installPath) continue;
+
+      // Get version
+      let version: string | null = null;
+      try {
+        const verResult = await sshExec(`${agent.name} --version`, { timeoutMs: 15_000 });
+        if (verResult.exitCode === 0) {
+          version = verResult.stdout.trim().slice(0, 100);
+        }
+      } catch {
+        // Some agents may not support --version — that's fine
+      }
+
+      // For ACP-capable agents, verify ACP mode actually works
+      let supportsAcp = agent.supportsAcp;
+      if (supportsAcp) {
+        try {
+          const acpCheck = await sshExec(`${agent.name} acp --help`, { timeoutMs: 10_000 });
+          supportsAcp = acpCheck.exitCode === 0;
+        } catch {
+          supportsAcp = false;
+        }
+      }
+
+      // UPSERT into available_agents
+      await sql`
+        INSERT INTO available_agents (name, install_path, version, supports_acp, last_probed_at)
+        VALUES (${agent.name}, ${installPath}, ${version}, ${supportsAcp}, clock_timestamp())
+        ON CONFLICT (name) DO UPDATE SET
+          install_path = EXCLUDED.install_path,
+          version = EXCLUDED.version,
+          supports_acp = EXCLUDED.supports_acp,
+          last_probed_at = EXCLUDED.last_probed_at
+      `;
+      log.info({ agent: agent.name, version, installPath, supportsAcp }, 'agent-probe: found on host');
+    } catch (err) {
+      // SSH failed or agent not found — skip silently
+      const msg = err instanceof Error ? err.message : String(err);
+      log.debug({ agent: agent.name, err: msg }, 'agent-probe: not found or SSH failed');
+    }
+  }
+
+  log.info('agent-probe: scan complete');
+}
--- a/apps/coder/src/services/dispatcher.ts
+++ b/apps/coder/src/services/dispatcher.ts
@@ -0,0 +1,384 @@
+import type { Sql } from '../db.js';
+import type { FastifyBaseLogger } from 'fastify';
+import type { Broker } from '@boocode/server/broker';
+import type { Config } from '../config.js';
+import { createWorktree, diffWorktree, cleanupWorktree } from './worktrees.js';
+import { dispatchViaAcp } from './acp-dispatch.js';
+import { dispatchViaPty } from './pty-dispatch.js';
+
+interface InferenceRunner {
+  enqueue: (sessionId: string, chatId: string, assistantId: string, user: string) => void;
+  cancel: (sessionId: string, chatId: string) => Promise<boolean>;
+  hasActive: (chatId: string) => boolean;
+}
+
+interface Deps {
+  sql: Sql;
+  inference: InferenceRunner;
+  broker: Broker;
+  log: FastifyBaseLogger;
+  config: Config;
+}
+
+const POLL_INTERVAL_MS = 5_000;
+const COMPLETION_POLL_MS = 2_000;
+
+export function createDispatcher(deps: Deps): { start(): void; stop(): Promise<void> } {
+  const { sql, inference, log, config } = deps;
+  let timer: ReturnType<typeof setInterval> | null = null;
+  let running = false;
+  let stopping = false;
+  let inflightPromise: Promise<void> | null = null;
+
+  async function poll(): Promise<void> {
+    if (running || stopping) return;
+
+    // Grab one pending task
+    const rows = await sql<{ id: string; project_id: string; input: string; agent: string | null; model: string | null }[]>`
+      SELECT id, project_id, input, agent, model
+      FROM tasks
+      WHERE state = 'pending'
+      ORDER BY created_at
+      LIMIT 1
+    `;
+    if (rows.length === 0) return;
+
+    const task = rows[0]!;
+    running = true;
+    inflightPromise = runTask(task).finally(() => {
+      running = false;
+      inflightPromise = null;
+    });
+  }
+
+  async function runTask(task: { id: string; project_id: string; input: string; agent: string | null; model: string | null }): Promise<void> {
+    const taskId = task.id;
+
+    // Determine execution path: if agent is specified AND exists in available_agents → Path B
+    if (task.agent) {
+      const [agentRow] = await sql<{ name: string; supports_acp: boolean }[]>`
+        SELECT name, supports_acp FROM available_agents WHERE name = ${task.agent}
+      `;
+      if (agentRow) {
+        await runExternalAgent(task, agentRow.supports_acp);
+        return;
+      }
+      // Agent specified but not available — fall through to Path A with a warning
+      log.warn({ taskId, agent: task.agent }, 'dispatcher: specified agent not available, falling back to native');
+    }
+
+    // Path A — native inference (existing behavior)
+    await runNativeInference(task);
+  }
+
+  // ─── Path A: Native Inference ───────────────────────────────────────────────
+
+  async function runNativeInference(task: { id: string; project_id: string; input: string; agent: string | null; model: string | null }): Promise<void> {
+    const taskId = task.id;
+    log.info({ taskId }, 'dispatcher: starting task (path A — native)');
+
+    try {
+      // Mark running
+      await sql`
+        UPDATE tasks
+        SET state = 'running', started_at = clock_timestamp(), execution_path = 'native'
+        WHERE id = ${taskId}
+      `;
+
+      // Create session + chat for this task
+      const model = task.model ?? config.DEFAULT_MODEL;
+      const sessionName = 'Task: ' + task.input.slice(0, 40);
+
+      const [session] = await sql<{ id: string }[]>`
+        INSERT INTO sessions (project_id, name, model, status)
+        VALUES (${task.project_id}, ${sessionName}, ${model}, 'open')
+        RETURNING id
+      `;
+      const sessionId = session!.id;
+
+      const [chat] = await sql<{ id: string }[]>`
+        INSERT INTO chats (session_id, name, status)
+        VALUES (${sessionId}, 'Task execution', 'open')
+        RETURNING id
+      `;
+      const chatId = chat!.id;
+
+      // Link task to session
+      await sql`UPDATE tasks SET session_id = ${sessionId} WHERE id = ${taskId}`;
+
+      // Create user message + streaming assistant
+      await sql<{ id: string }[]>`
+        INSERT INTO messages (session_id, chat_id, role, content, status, created_at)
+        VALUES (${sessionId}, ${chatId}, 'user', ${task.input}, 'complete', clock_timestamp())
+        RETURNING id
+      `;
+      const [assistantMsg] = await sql<{ id: string }[]>`
+        INSERT INTO messages (session_id, chat_id, role, content, status, created_at)
+        VALUES (${sessionId}, ${chatId}, 'assistant', '', 'streaming', clock_timestamp())
+        RETURNING id
+      `;
+      const assistantId = assistantMsg!.id;
+
+      // Enqueue inference
+      inference.enqueue(sessionId, chatId, assistantId, 'default');
+
+      // Wait for inference to complete (poll message status)
+      const finalStatus = await waitForCompletion(assistantId);
+
+      if (stopping) {
+        await sql`
+          UPDATE tasks
+          SET state = 'cancelled', ended_at = clock_timestamp()
+          WHERE id = ${taskId}
+        `;
+        return;
+      }
+
+      // Aggregate token cost for the task's session
+      const [costRow] = await sql<{ total: number | null }[]>`
+        SELECT SUM(tokens_used)::int AS total
+        FROM messages
+        WHERE session_id = ${sessionId} AND tokens_used IS NOT NULL
+      `;
+      const costTokens = costRow?.total ?? null;
+
+      if (finalStatus === 'complete') {
+        const [msg] = await sql<{ content: string | null }[]>`
+          SELECT content FROM messages WHERE id = ${assistantId}
+        `;
+        const summary = (msg?.content ?? '').slice(0, 500);
+        await sql`
+          UPDATE tasks
+          SET state = 'completed', ended_at = clock_timestamp(), output_summary = ${summary}, cost_tokens = ${costTokens}
+          WHERE id = ${taskId}
+        `;
+        log.info({ taskId, costTokens }, 'dispatcher: task completed (native)');
+      } else {
+        const [msg] = await sql<{ content: string | null }[]>`
+          SELECT content FROM messages WHERE id = ${assistantId}
+        `;
+        const summary = (msg?.content ?? 'Inference failed').slice(0, 500);
+        await sql`
+          UPDATE tasks
+          SET state = 'failed', ended_at = clock_timestamp(), output_summary = ${summary}, cost_tokens = ${costTokens}
+          WHERE id = ${taskId}
+        `;
+        log.warn({ taskId, finalStatus }, 'dispatcher: task failed (native)');
+      }
+    } catch (err) {
+      const errMsg = err instanceof Error ? err.message : String(err);
+      log.error({ taskId, err: errMsg }, 'dispatcher: task error (native)');
+      await sql`
+        UPDATE tasks
+        SET state = 'failed', ended_at = clock_timestamp(), output_summary = ${errMsg.slice(0, 500)}
+        WHERE id = ${taskId}
+      `.catch(() => {});
+    }
+  }
+
+  // ─── Path B: External Agent Dispatch ──────<E29480><E29480><EFBFBD>─────────────────────────────────
+
+  async function runExternalAgent(
+    task: { id: string; project_id: string; input: string; agent: string | null; model: string | null },
+    supportsAcp: boolean,
+  ): Promise<void> {
+    const taskId = task.id;
+    const agent = task.agent!;
+    const executionPath = supportsAcp ? 'acp' : 'pty';
+
+    log.info({ taskId, agent, executionPath }, 'dispatcher: starting task (path B — external)');
+
+    // Resolve the project's root path
+    const [project] = await sql<{ root_path: string | null }[]>`
+      SELECT root_path FROM projects WHERE id = ${task.project_id}
+    `;
+    const projectPath = project?.root_path;
+    if (!projectPath) {
+      await sql`
+        UPDATE tasks
+        SET state = 'failed', ended_at = clock_timestamp(), output_summary = 'Project has no root_path — cannot create worktree'
+        WHERE id = ${taskId}
+      `;
+      return;
+    }
+
+    // Create an abort controller for this task
+    const ac = new AbortController();
+
+    try {
+      // Mark running
+      await sql`
+        UPDATE tasks
+        SET state = 'running', started_at = clock_timestamp(), execution_path = ${executionPath}
+        WHERE id = ${taskId}
+      `;
+
+      // Create session + chat for this task (same as Path A — for output tracking)
+      const sessionName = `Task [${agent}]: ${task.input.slice(0, 30)}`;
+      const [session] = await sql<{ id: string }[]>`
+        INSERT INTO sessions (project_id, name, model, status)
+        VALUES (${task.project_id}, ${sessionName}, ${task.model ?? config.DEFAULT_MODEL}, 'open')
+        RETURNING id
+      `;
+      const sessionId = session!.id;
+
+      const [chat] = await sql<{ id: string }[]>`
+        INSERT INTO chats (session_id, name, status)
+        VALUES (${sessionId}, 'External agent execution', 'open')
+        RETURNING id
+      `;
+      const chatId = chat!.id;
+
+      // Link task to session
+      await sql`UPDATE tasks SET session_id = ${sessionId} WHERE id = ${taskId}`;
+
+      // Create user message for the task input
+      await sql`
+        INSERT INTO messages (session_id, chat_id, role, content, status, created_at)
+        VALUES (${sessionId}, ${chatId}, 'user', ${task.input}, 'complete', clock_timestamp())
+      `;
+
+      // Step 1: Create worktree
+      log.info({ taskId, projectPath }, 'dispatcher: creating worktree');
+      const worktreePath = await createWorktree(projectPath, taskId, { signal: ac.signal });
+      log.info({ taskId, worktreePath }, 'dispatcher: worktree created');
+
+      // Step 2: Dispatch to agent
+      let outputSummary: string;
+
+      if (supportsAcp) {
+        const result = await dispatchViaAcp({
+          agent,
+          task: task.input,
+          worktreePath,
+          model: task.model ?? undefined,
+          signal: ac.signal,
+          log,
+        });
+        outputSummary = result.output.slice(0, 500);
+
+        // Store agent output as an assistant message
+        await sql`
+          INSERT INTO messages (session_id, chat_id, role, content, status, created_at)
+          VALUES (${sessionId}, ${chatId}, 'assistant', ${result.output.slice(0, 50_000)}, 'complete', clock_timestamp())
+        `;
+      } else {
+        const result = await dispatchViaPty({
+          agent,
+          task: task.input,
+          worktreePath,
+          model: task.model ?? undefined,
+          signal: ac.signal,
+          log,
+        });
+        outputSummary = (result.stdout || result.stderr).slice(0, 500);
+
+        // Store agent output as an assistant message
+        const content = result.stdout || result.stderr || '(no output)';
+        await sql`
+          INSERT INTO messages (session_id, chat_id, role, content, status, created_at)
+          VALUES (${sessionId}, ${chatId}, 'assistant', ${content.slice(0, 50_000)}, 'complete', clock_timestamp())
+        `;
+      }
+
+      if (stopping) {
+        await sql`
+          UPDATE tasks SET state = 'cancelled', ended_at = clock_timestamp() WHERE id = ${taskId}
+        `;
+        await cleanupWorktree(projectPath, taskId);
+        return;
+      }
+
+      // Step 3: Diff the worktree and queue pending changes
+      log.info({ taskId }, 'dispatcher: diffing worktree');
+      const diff = await diffWorktree(worktreePath, projectPath, { signal: ac.signal });
+
+      if (diff) {
+        // Queue a single pending_change entry with the full unified diff
+        await sql`
+          INSERT INTO pending_changes (session_id, task_id, file_path, operation, diff)
+          VALUES (${sessionId}, ${taskId}, ${projectPath}, 'edit', ${diff})
+        `;
+        log.info({ taskId, diffLength: diff.length }, 'dispatcher: diff queued as pending change');
+      } else {
+        log.info({ taskId }, 'dispatcher: no changes detected in worktree');
+      }
+
+      // Step 4: Cleanup worktree
+      await cleanupWorktree(projectPath, taskId);
+
+      // Step 5: Aggregate token cost
+      const [extCostRow] = await sql<{ total: number | null }[]>`
+        SELECT SUM(tokens_used)::int AS total
+        FROM messages
+        WHERE session_id = ${sessionId} AND tokens_used IS NOT NULL
+      `;
+      const extCostTokens = extCostRow?.total ?? null;
+
+      // Step 6: Mark task completed
+      await sql`
+        UPDATE tasks
+        SET state = 'completed', ended_at = clock_timestamp(), output_summary = ${outputSummary}, cost_tokens = ${extCostTokens}
+        WHERE id = ${taskId}
+      `;
+      log.info({ taskId, agent, costTokens: extCostTokens }, 'dispatcher: task completed (external)');
+
+    } catch (err) {
+      const errMsg = err instanceof Error ? err.message : String(err);
+      log.error({ taskId, agent, err: errMsg }, 'dispatcher: external agent error');
+
+      await sql`
+        UPDATE tasks
+        SET state = 'failed', ended_at = clock_timestamp(), output_summary = ${errMsg.slice(0, 500)}
+        WHERE id = ${taskId}
+      `.catch(() => {});
+
+      // Best-effort cleanup
+      await cleanupWorktree(projectPath, taskId);
+    }
+  }
+
+  // ─── Helpers ────────────────────────────────────────────────────────────────
+
+  async function waitForCompletion(assistantId: string): Promise<string> {
+    for (;;) {
+      if (stopping) return 'cancelled';
+
+      const [row] = await sql<{ status: string }[]>`
+        SELECT status FROM messages WHERE id = ${assistantId}
+      `;
+      const status = row?.status ?? 'failed';
+      if (status !== 'streaming') return status;
+
+      await sleep(COMPLETION_POLL_MS);
+    }
+  }
+
+  function sleep(ms: number): Promise<void> {
+    return new Promise((resolve) => setTimeout(resolve, ms));
+  }
+
+  return {
+    start() {
+      log.info('dispatcher: starting poll loop');
+      timer = setInterval(() => {
+        poll().catch((err) => {
+          log.error({ err }, 'dispatcher: poll error');
+        });
+      }, POLL_INTERVAL_MS);
+    },
+
+    async stop() {
+      stopping = true;
+      if (timer) {
+        clearInterval(timer);
+        timer = null;
+      }
+      if (inflightPromise) {
+        log.info('dispatcher: waiting for in-flight task');
+        await inflightPromise;
+      }
+      log.info('dispatcher: stopped');
+    },
+  };
+}
--- a/apps/coder/src/services/mcp-server.ts
+++ b/apps/coder/src/services/mcp-server.ts
@@ -0,0 +1,201 @@
+/**
+ * BooCoder MCP Server — exposes task primitives as MCP tools.
+ *
+ * Started when `--mcp` flag is passed to the entry point. Runs stdio transport
+ * so external tools (opencode in Termius) can drive the task queue.
+ */
+import { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js';
+import { StdioServerTransport } from '@modelcontextprotocol/sdk/server/stdio.js';
+import { z } from 'zod';
+import type { Sql } from '../db.js';
+import { applyOne, rejectOne } from './pending_changes.js';
+
+// --- Tool handlers -----------------------------------------------------------
+
+interface TaskRow {
+  id: string;
+  state: string;
+}
+
+interface PendingRow {
+  id: string;
+  file_path: string;
+  operation: string;
+  diff: string;
+  session_id: string;
+}
+
+interface WorktreeRow {
+  id: string;
+  worktree_path: string;
+  agent: string;
+  started_at: string;
+}
+
+interface ProjectPathRow {
+  path: string;
+}
+
+function textResult(data: unknown) {
+  return { content: [{ type: 'text' as const, text: JSON.stringify(data, null, 2) }] };
+}
+
+// --- Public entry ------------------------------------------------------------
+
+export async function startMcpServer(sql: Sql): Promise<void> {
+  const server = new McpServer(
+    { name: 'boocoder', version: '2.0.2' },
+    { capabilities: { tools: {} } },
+  );
+
+  // 1. boocoder.create_task
+  server.tool(
+    'boocoder.create_task',
+    'Create a new task in the BooCoder task queue',
+    {
+      project_id: z.string().describe('Project UUID'),
+      input: z.string().describe('Task description / prompt for the agent'),
+      agent: z.string().optional().describe('Agent name (optional — uses default if omitted)'),
+      model: z.string().optional().describe('Model override (optional)'),
+    },
+    async (args) => {
+      const [row] = await sql<TaskRow[]>`
+        INSERT INTO tasks (project_id, input, agent, model, state)
+        VALUES (${args.project_id}, ${args.input}, ${args.agent ?? null}, ${args.model ?? null}, 'pending')
+        RETURNING id, state
+      `;
+      return textResult({ task_id: row!.id, state: row!.state });
+    },
+  );
+
+  // 2. boocoder.list_pending_changes
+  server.tool(
+    'boocoder.list_pending_changes',
+    'List pending changes awaiting review',
+    {
+      session_id: z.string().optional().describe('Optional session filter'),
+    },
+    async (args) => {
+      let rows: PendingRow[];
+      if (args.session_id) {
+        rows = await sql<PendingRow[]>`
+          SELECT id, file_path, operation, diff, session_id
+          FROM pending_changes
+          WHERE status = 'pending' AND session_id = ${args.session_id}
+          ORDER BY created_at ASC
+        `;
+      } else {
+        rows = await sql<PendingRow[]>`
+          SELECT id, file_path, operation, diff, session_id
+          FROM pending_changes
+          WHERE status = 'pending'
+          ORDER BY created_at ASC
+        `;
+      }
+      const items = rows.map((r) => ({
+        id: r.id,
+        file_path: r.file_path,
+        operation: r.operation,
+        diff_preview: r.diff.slice(0, 200),
+      }));
+      return textResult(items);
+    },
+  );
+
+  // 3. boocoder.apply
+  server.tool(
+    'boocoder.apply',
+    'Apply a pending change (write to disk)',
+    {
+      change_id: z.string().describe('Pending change UUID'),
+    },
+    async (args) => {
+      // Resolve projectRoot from the change's session → project path
+      const [proj] = await sql<ProjectPathRow[]>`
+        SELECT p.path FROM pending_changes pc
+        JOIN sessions s ON pc.session_id = s.id
+        JOIN projects p ON s.project_id = p.id
+        WHERE pc.id = ${args.change_id}
+      `;
+      if (!proj) {
+        return textResult({ success: false, file_path: '', error: 'change not found or project path unresolved' });
+      }
+      const result = await applyOne(sql, args.change_id, proj.path);
+      return textResult({ success: result.success, file_path: result.file_path, error: result.error });
+    },
+  );
+
+  // 4. boocoder.reject
+  server.tool(
+    'boocoder.reject',
+    'Reject a pending change (mark as rejected, no disk write)',
+    {
+      change_id: z.string().describe('Pending change UUID'),
+    },
+    async (args) => {
+      await rejectOne(sql, args.change_id);
+      return textResult({ success: true });
+    },
+  );
+
+  // 5. boocoder.dispatch_external_agent
+  server.tool(
+    'boocoder.dispatch_external_agent',
+    'Create a task targeting a specific external agent (ACP or PTY dispatch)',
+    {
+      project_id: z.string().describe('Project UUID'),
+      input: z.string().describe('Task prompt'),
+      agent: z.string().describe('Agent name (must match available_agents registry)'),
+      model: z.string().optional().describe('Model override (optional)'),
+    },
+    async (args) => {
+      const [row] = await sql<TaskRow[]>`
+        INSERT INTO tasks (project_id, input, agent, model, state)
+        VALUES (${args.project_id}, ${args.input}, ${args.agent}, ${args.model ?? null}, 'pending')
+        RETURNING id, state
+      `;
+
+      // Determine execution path from available_agents
+      const [agentRow] = await sql<{ supports_acp: boolean }[]>`
+        SELECT supports_acp FROM available_agents WHERE name = ${args.agent}
+      `;
+      const executionPath = agentRow?.supports_acp ? 'acp' : 'pty';
+
+      return textResult({ task_id: row!.id, state: row!.state, execution_path: executionPath });
+    },
+  );
+
+  // 6. boocoder.list_worktrees
+  server.tool(
+    'boocoder.list_worktrees',
+    'List active worktrees from running tasks',
+    {},
+    async () => {
+      const rows = await sql<WorktreeRow[]>`
+        SELECT id, worktree_path, agent, started_at
+        FROM tasks
+        WHERE worktree_path IS NOT NULL AND state = 'running'
+        ORDER BY started_at DESC
+      `;
+      const items = rows.map((r) => ({
+        task_id: r.id,
+        worktree_path: r.worktree_path,
+        agent: r.agent,
+        started_at: r.started_at,
+      }));
+      return textResult(items);
+    },
+  );
+
+  // Connect via stdio
+  const transport = new StdioServerTransport();
+  await server.connect(transport);
+
+  // Block until stdin closes (transport handles lifecycle)
+  await new Promise<void>((resolve) => {
+    process.stdin.on('end', resolve);
+    process.stdin.on('close', resolve);
+  });
+
+  await sql.end({ timeout: 5 });
+}
--- a/apps/coder/src/services/pending_changes.ts
+++ b/apps/coder/src/services/pending_changes.ts
@@ -0,0 +1,224 @@
+import { readFile, writeFile, unlink, mkdir } from 'node:fs/promises';
+import { dirname } from 'node:path';
+import type { Sql } from '../db.js';
+import { resolveWritePath } from './write_guard.js';
+
+// --- Types -------------------------------------------------------------------
+
+export interface PendingChange {
+  id: string;
+  session_id: string;
+  task_id: string | null;
+  file_path: string;
+  operation: 'create' | 'edit' | 'delete';
+  diff: string;
+  status: 'pending' | 'applied' | 'rejected' | 'reverted';
+  created_at: string;
+}
+
+export interface ApplyResult {
+  id: string;
+  file_path: string;
+  operation: string;
+  success: boolean;
+  error?: string;
+}
+
+// --- Queue functions ---------------------------------------------------------
+
+export async function queueEdit(
+  sql: Sql,
+  sessionId: string,
+  taskId: string | null,
+  filePath: string,
+  oldString: string,
+  newString: string,
+  projectRoot: string,
+): Promise<PendingChange> {
+  const resolved = resolveWritePath(projectRoot, filePath);
+  const diff = JSON.stringify({ old: oldString, new: newString });
+
+  const [row] = await sql<PendingChange[]>`
+    INSERT INTO pending_changes (session_id, task_id, file_path, operation, diff)
+    VALUES (${sessionId}, ${taskId}, ${resolved}, 'edit', ${diff})
+    RETURNING *
+  `;
+  return row!;
+}
+
+export async function queueCreate(
+  sql: Sql,
+  sessionId: string,
+  taskId: string | null,
+  filePath: string,
+  content: string,
+  projectRoot: string,
+): Promise<PendingChange> {
+  const resolved = resolveWritePath(projectRoot, filePath);
+
+  const [row] = await sql<PendingChange[]>`
+    INSERT INTO pending_changes (session_id, task_id, file_path, operation, diff)
+    VALUES (${sessionId}, ${taskId}, ${resolved}, 'create', ${content})
+    RETURNING *
+  `;
+  return row!;
+}
+
+export async function queueDelete(
+  sql: Sql,
+  sessionId: string,
+  taskId: string | null,
+  filePath: string,
+  projectRoot: string,
+): Promise<PendingChange> {
+  const resolved = resolveWritePath(projectRoot, filePath);
+
+  const [row] = await sql<PendingChange[]>`
+    INSERT INTO pending_changes (session_id, task_id, file_path, operation, diff)
+    VALUES (${sessionId}, ${taskId}, ${resolved}, 'delete', '')
+    RETURNING *
+  `;
+  return row!;
+}
+
+// --- Apply functions ---------------------------------------------------------
+
+export async function applyOne(
+  sql: Sql,
+  changeId: string,
+  projectRoot: string,
+): Promise<ApplyResult> {
+  const [change] = await sql<PendingChange[]>`
+    SELECT * FROM pending_changes WHERE id = ${changeId} AND status = 'pending'
+  `;
+  if (!change) {
+    return { id: changeId, file_path: '', operation: '', success: false, error: 'change not found or not pending' };
+  }
+
+  try {
+    // Re-validate path in case projectRoot has shifted
+    resolveWritePath(projectRoot, change.file_path);
+
+    switch (change.operation) {
+      case 'create': {
+        await mkdir(dirname(change.file_path), { recursive: true });
+        await writeFile(change.file_path, change.diff, 'utf8');
+        break;
+      }
+      case 'edit': {
+        const { old: oldStr, new: newStr } = JSON.parse(change.diff) as { old: string; new: string };
+        const content = await readFile(change.file_path, 'utf8');
+        if (!content.includes(oldStr)) {
+          throw new Error('old_string not found in file — file may have changed since the edit was queued');
+        }
+        const updated = content.replace(oldStr, newStr);
+        await writeFile(change.file_path, updated, 'utf8');
+        break;
+      }
+      case 'delete': {
+        // Stash current content in diff for potential rewind
+        try {
+          const existing = await readFile(change.file_path, 'utf8');
+          await sql`UPDATE pending_changes SET diff = ${existing} WHERE id = ${changeId}`;
+        } catch {
+          // File may already be gone — proceed with status update
+        }
+        await unlink(change.file_path);
+        break;
+      }
+    }
+
+    await sql`UPDATE pending_changes SET status = 'applied' WHERE id = ${changeId}`;
+    return { id: change.id, file_path: change.file_path, operation: change.operation, success: true };
+  } catch (err) {
+    const message = err instanceof Error ? err.message : String(err);
+    return { id: change.id, file_path: change.file_path, operation: change.operation, success: false, error: message };
+  }
+}
+
+export async function applyAll(
+  sql: Sql,
+  sessionId: string,
+  projectRoot: string,
+): Promise<ApplyResult[]> {
+  const pending = await sql<PendingChange[]>`
+    SELECT * FROM pending_changes
+    WHERE session_id = ${sessionId} AND status = 'pending'
+    ORDER BY created_at ASC
+  `;
+  const results: ApplyResult[] = [];
+  for (const change of pending) {
+    results.push(await applyOne(sql, change.id, projectRoot));
+  }
+  return results;
+}
+
+// --- Reject functions --------------------------------------------------------
+
+export async function rejectOne(sql: Sql, changeId: string): Promise<void> {
+  await sql`UPDATE pending_changes SET status = 'rejected' WHERE id = ${changeId} AND status = 'pending'`;
+}
+
+export async function rejectAll(sql: Sql, sessionId: string): Promise<void> {
+  await sql`UPDATE pending_changes SET status = 'rejected' WHERE session_id = ${sessionId} AND status = 'pending'`;
+}
+
+// --- Rewind functions --------------------------------------------------------
+
+export async function rewindOne(
+  sql: Sql,
+  changeId: string,
+  projectRoot: string,
+): Promise<ApplyResult> {
+  const [change] = await sql<PendingChange[]>`
+    SELECT * FROM pending_changes WHERE id = ${changeId} AND status = 'applied'
+  `;
+  if (!change) {
+    return { id: changeId, file_path: '', operation: '', success: false, error: 'change not found or not applied' };
+  }
+
+  try {
+    resolveWritePath(projectRoot, change.file_path);
+
+    switch (change.operation) {
+      case 'create': {
+        // Reverse a create: delete the file
+        await unlink(change.file_path);
+        break;
+      }
+      case 'edit': {
+        // Reverse an edit: swap old and new
+        const { old: oldStr, new: newStr } = JSON.parse(change.diff) as { old: string; new: string };
+        const content = await readFile(change.file_path, 'utf8');
+        if (!content.includes(newStr)) {
+          throw new Error('new_string not found in file — cannot rewind; file may have been modified since apply');
+        }
+        const reverted = content.replace(newStr, oldStr);
+        await writeFile(change.file_path, reverted, 'utf8');
+        break;
+      }
+      case 'delete': {
+        // Reverse a delete: recreate the file (diff holds the original content stashed at apply time)
+        await mkdir(dirname(change.file_path), { recursive: true });
+        await writeFile(change.file_path, change.diff, 'utf8');
+        break;
+      }
+    }
+
+    await sql`UPDATE pending_changes SET status = 'reverted' WHERE id = ${changeId}`;
+    return { id: change.id, file_path: change.file_path, operation: change.operation, success: true };
+  } catch (err) {
+    const message = err instanceof Error ? err.message : String(err);
+    return { id: change.id, file_path: change.file_path, operation: change.operation, success: false, error: message };
+  }
+}
+
+// --- Query functions ---------------------------------------------------------
+
+export async function listPending(sql: Sql, sessionId: string): Promise<PendingChange[]> {
+  return sql<PendingChange[]>`
+    SELECT * FROM pending_changes
+    WHERE session_id = ${sessionId} AND status = 'pending'
+    ORDER BY created_at ASC
+  `;
+}
--- a/apps/coder/src/services/pty-dispatch.ts
+++ b/apps/coder/src/services/pty-dispatch.ts
@@ -0,0 +1,146 @@
+/**
+ * PTY dispatch — runs external agents on the host via SSH.
+ *
+ * For agents without ACP support (claude, pi), we pipe the task into their
+ * non-interactive mode and capture stdout/stderr. The agent runs in a git
+ * worktree so it can modify files freely.
+ *
+ * Supported agents:
+ * - claude: `claude -p --model <model>` (print mode, reads task from stdin)
+ * - opencode: `echo <task> | opencode` (stdin pipe — exact flags TBD)
+ * - qwen: `qwen -p <task> --output-format stream-json` (NDJSON structured output)
+ * - goose: stub (not yet supported)
+ * - pi: stub (not yet supported)
+ */
+import type { FastifyBaseLogger } from 'fastify';
+import { sshSpawnWithStdin } from './ssh.js';
+
+export interface DispatchResult {
+  exitCode: number;
+  stdout: string;
+  stderr: string;
+}
+
+export interface PtyDispatchOpts {
+  agent: string;
+  task: string;
+  worktreePath: string;
+  model?: string;
+  signal?: AbortSignal;
+  log: FastifyBaseLogger;
+}
+
+/**
+ * Build the shell command that runs the agent non-interactively.
+ * The command will be executed inside `cd <worktreePath> && ...`.
+ */
+function buildAgentCommand(agent: string, task: string, model?: string): string | null {
+  // Escape the task for embedding in a shell command
+  const escapedTask = task.replace(/'/g, "'\\''");
+
+  switch (agent) {
+    case 'claude':
+      // Claude Code's print mode: reads prompt from stdin, runs autonomously, prints result
+      return model
+        ? `echo '${escapedTask}' | claude -p --model '${model}'`
+        : `echo '${escapedTask}' | claude -p`;
+
+    case 'opencode':
+      // opencode non-interactive: pipe task via stdin
+      // NOTE: exact flags may vary — opencode may need --non-interactive or --pipe
+      return model
+        ? `echo '${escapedTask}' | opencode --model '${model}'`
+        : `echo '${escapedTask}' | opencode`;
+
+    case 'qwen':
+      // Qwen Code: structured JSON output mode for parseable events
+      return model
+        ? `qwen -p '${escapedTask}' --model '${model}' --output-format stream-json`
+        : `qwen -p '${escapedTask}' --output-format stream-json`;
+
+    case 'goose':
+      // Not yet verified for non-interactive use
+      return null;
+
+    case 'pi':
+      // Not yet verified for non-interactive use
+      return null;
+
+    default:
+      return null;
+  }
+}
+
+/**
+ * Dispatch a task to an external agent via SSH.
+ *
+ * The agent runs in the worktree directory on the host. stdout/stderr are
+ * captured in full and returned. The SSH process is killed on abort signal.
+ */
+export async function dispatchViaPty(opts: PtyDispatchOpts): Promise<DispatchResult> {
+  const { agent, task, worktreePath, model, signal, log } = opts;
+
+  const agentCmd = buildAgentCommand(agent, task, model);
+  if (!agentCmd) {
+    return {
+      exitCode: 1,
+      stdout: '',
+      stderr: `Agent '${agent}' is not yet supported for PTY dispatch.`,
+    };
+  }
+
+  // Wrap in cd to the worktree
+  const fullCommand = `cd '${worktreePath.replace(/'/g, "'\\''")}' && ${agentCmd}`;
+
+  log.info({ agent, worktreePath }, 'pty-dispatch: starting');
+
+  return new Promise<DispatchResult>((resolve, reject) => {
+    const child = sshSpawnWithStdin(fullCommand, '');
+    // Note: sshSpawnWithStdin already closes stdin. For agents that read from
+    // stdin via echo piping, the command itself handles the piping on the remote
+    // side. We just need the SSH tunnel.
+
+    // Actually, re-think: sshSpawnWithStdin writes input and closes stdin on the
+    // LOCAL ssh process. But the remote command is `echo '...' | agent`, which
+    // provides its own stdin. So we should use sshSpawn (no local stdin needed)
+    // or just let the empty stdin close — the remote shell handles piping internally.
+    // This is fine as-is because the echo piping happens WITHIN the remote shell command.
+
+    let stdout = '';
+    let stderr = '';
+    let killed = false;
+
+    child.stdout!.on('data', (chunk: Buffer) => { stdout += chunk.toString(); });
+    child.stderr!.on('data', (chunk: Buffer) => { stderr += chunk.toString(); });
+
+    const cleanup = () => {
+      if (!killed) {
+        killed = true;
+        child.kill('SIGTERM');
+        // Give it a moment then force-kill
+        setTimeout(() => child.kill('SIGKILL'), 5_000);
+      }
+    };
+
+    if (signal) {
+      if (signal.aborted) {
+        cleanup();
+        resolve({ exitCode: 130, stdout: '', stderr: 'Aborted before start' });
+        return;
+      }
+      signal.addEventListener('abort', cleanup, { once: true });
+    }
+
+    child.on('close', (code) => {
+      if (signal) signal.removeEventListener('abort', cleanup);
+      log.info({ agent, exitCode: code }, 'pty-dispatch: completed');
+      resolve({ exitCode: code ?? 1, stdout, stderr });
+    });
+
+    child.on('error', (err) => {
+      if (signal) signal.removeEventListener('abort', cleanup);
+      log.error({ agent, err: err.message }, 'pty-dispatch: spawn error');
+      reject(err);
+    });
+  });
+}
--- a/apps/coder/src/services/ssh.ts
+++ b/apps/coder/src/services/ssh.ts
@@ -0,0 +1,126 @@
+/**
+ * SSH helper — spawns commands on the host via SSH.
+ *
+ * BooCode's container cannot directly spawn host processes (opencode, goose, claude, pi).
+ * They live on the HOST at /usr/local/bin/ or Sam's PATH. We SSH to the host over the
+ * Tailscale IP (same mechanism BooTerm uses: samkintop@100.114.205.53).
+ */
+import { spawn, type ChildProcess } from 'node:child_process';
+
+export const SSH_HOST = process.env.BOOCODER_SSH_HOST ?? '100.114.205.53';
+export const SSH_USER = process.env.BOOCODER_SSH_USER ?? 'samkintop';
+
+/** Common SSH args — strict host checking disabled for container-to-host trust. */
+const SSH_BASE_ARGS = [
+  '-o', 'StrictHostKeyChecking=no',
+  '-o', 'UserKnownHostsFile=/dev/null',
+  '-o', 'LogLevel=ERROR',
+  '-o', 'BatchMode=yes',
+];
+
+export interface SshExecResult {
+  exitCode: number;
+  stdout: string;
+  stderr: string;
+}
+
+/**
+ * Execute a command on the host via SSH, collecting all output.
+ * Returns when the remote process exits.
+ */
+export async function sshExec(
+  command: string,
+  opts?: { signal?: AbortSignal; timeoutMs?: number },
+): Promise<SshExecResult> {
+  return new Promise<SshExecResult>((resolve, reject) => {
+    const child = spawn('ssh', [
+      ...SSH_BASE_ARGS,
+      `${SSH_USER}@${SSH_HOST}`,
+      command,
+    ], {
+      stdio: ['pipe', 'pipe', 'pipe'],
+    });
+
+    let stdout = '';
+    let stderr = '';
+    let killed = false;
+
+    child.stdout!.on('data', (chunk: Buffer) => { stdout += chunk.toString(); });
+    child.stderr!.on('data', (chunk: Buffer) => { stderr += chunk.toString(); });
+
+    const cleanup = () => {
+      if (!killed) {
+        killed = true;
+        child.kill('SIGTERM');
+      }
+    };
+
+    // Abort signal
+    if (opts?.signal) {
+      if (opts.signal.aborted) {
+        cleanup();
+        reject(new Error('SSH exec aborted before start'));
+        return;
+      }
+      opts.signal.addEventListener('abort', cleanup, { once: true });
+    }
+
+    // Timeout
+    let timer: ReturnType<typeof setTimeout> | undefined;
+    if (opts?.timeoutMs) {
+      timer = setTimeout(() => {
+        cleanup();
+        reject(new Error(`SSH exec timed out after ${opts.timeoutMs}ms`));
+      }, opts.timeoutMs);
+    }
+
+    child.on('close', (code) => {
+      if (timer) clearTimeout(timer);
+      if (opts?.signal) opts.signal.removeEventListener('abort', cleanup);
+      resolve({ exitCode: code ?? 1, stdout, stderr });
+    });
+
+    child.on('error', (err) => {
+      if (timer) clearTimeout(timer);
+      if (opts?.signal) opts.signal.removeEventListener('abort', cleanup);
+      reject(err);
+    });
+
+    // Close stdin immediately — we're not sending input via sshExec
+    child.stdin!.end();
+  });
+}
+
+/**
+ * Spawn an SSH child process with a command on the host.
+ * Returns the raw ChildProcess for callers that need streaming I/O (ACP, PTY).
+ */
+export function sshSpawn(command: string): ChildProcess {
+  return spawn('ssh', [
+    ...SSH_BASE_ARGS,
+    `${SSH_USER}@${SSH_HOST}`,
+    command,
+  ], {
+    stdio: ['pipe', 'pipe', 'pipe'],
+  });
+}
+
+/**
+ * Spawn an SSH child process that pipes stdin through.
+ * Used for agents that read a task from stdin (e.g. `echo "task" | claude -p`).
+ */
+export function sshSpawnWithStdin(command: string, input: string): ChildProcess {
+  const child = spawn('ssh', [
+    ...SSH_BASE_ARGS,
+    `${SSH_USER}@${SSH_HOST}`,
+    command,
+  ], {
+    stdio: ['pipe', 'pipe', 'pipe'],
+  });
+
+  // Write the input and close stdin
+  child.stdin!.write(input);
+  child.stdin!.end();
+
+  return child;
+}
--- a/apps/coder/src/services/tools/adapter.ts
+++ b/apps/coder/src/services/tools/adapter.ts
@@ -0,0 +1,30 @@
+/**
+ * Adapts BooCoder write tools (which take ToolContext) into BooChat's ToolDef
+ * interface (which takes `projectRoot, extraRoots?`).
+ *
+ * The adapter reads the module-level inference context at execute time, so the
+ * wrapping happens at boot (static) — no per-inference re-wrap needed.
+ */
+
+import type { ToolDef as ServerToolDef } from '@boocode/server/tools';
+import type { ToolDef, ToolContext } from './types.js';
+import { getInferenceContext } from './inference_context.js';
+
+/**
+ * Wrap a BooCoder write tool (execute takes ToolContext) into a BooChat
+ * ToolDef (execute takes projectRoot + optional extraRoots). The adapter
+ * builds the ToolContext from the module-level inference context at call time.
+ */
+// eslint-disable-next-line @typescript-eslint/no-explicit-any
+export function adaptWriteTool(tool: ToolDef<any>): ServerToolDef<any> {
+  return {
+    name: tool.name,
+    description: tool.description,
+    inputSchema: tool.inputSchema,
+    jsonSchema: tool.jsonSchema,
+    async execute(input: unknown, projectRoot: string, _extraRoots?: readonly string[]): Promise<unknown> {
+      const ctx: ToolContext = getInferenceContext();
+      return tool.execute(input, projectRoot, ctx);
+    },
+  };
+}
--- a/apps/coder/src/services/tools/apply_pending.ts
+++ b/apps/coder/src/services/tools/apply_pending.ts
@@ -0,0 +1,44 @@
+import { z } from 'zod';
+import type { ToolDef, ToolContext } from './types.js';
+import { applyAll } from '../pending_changes.js';
+
+const ApplyPendingInput = z.object({});
+type ApplyPendingInputT = z.infer<typeof ApplyPendingInput>;
+
+export const applyPendingTool: ToolDef<ApplyPendingInputT> = {
+  name: 'apply_pending',
+  description:
+    'Apply all pending changes for the current session to disk. ' +
+    'Each queued create/edit/delete is executed in order.',
+  inputSchema: ApplyPendingInput,
+  jsonSchema: {
+    type: 'function',
+    function: {
+      name: 'apply_pending',
+      description:
+        'Apply all pending changes for the current session to disk. ' +
+        'Each queued create/edit/delete is executed in order.',
+      parameters: {
+        type: 'object',
+        properties: {},
+        required: [],
+      },
+    },
+  },
+  async execute(_input: ApplyPendingInputT, projectRoot: string, context: ToolContext): Promise<unknown> {
+    const results = await applyAll(context.sql, context.sessionId, projectRoot);
+    const succeeded = results.filter((r) => r.success).length;
+    const failed = results.filter((r) => !r.success).length;
+
+    return {
+      total: results.length,
+      succeeded,
+      failed,
+      results,
+      message:
+        results.length === 0
+          ? 'No pending changes to apply.'
+          : `Applied ${succeeded}/${results.length} changes.${failed > 0 ? ` ${failed} failed.` : ''}`,
+    };
+  },
+};
--- a/apps/coder/src/services/tools/check_task_status.ts
+++ b/apps/coder/src/services/tools/check_task_status.ts
@@ -0,0 +1,50 @@
+import { z } from 'zod';
+import type { ToolDef, ToolContext } from './types.js';
+
+const CheckTaskStatusInput = z.object({
+  task_id: z.string().uuid().describe('ID of the task to check'),
+});
+
+type CheckTaskStatusInputT = z.infer<typeof CheckTaskStatusInput>;
+
+export const checkTaskStatusTool: ToolDef<CheckTaskStatusInputT> = {
+  name: 'check_task_status',
+  description: 'Check the status and output of a subtask by ID. Returns state, output_summary, and timing.',
+  inputSchema: CheckTaskStatusInput,
+  jsonSchema: {
+    type: 'function',
+    function: {
+      name: 'check_task_status',
+      description: 'Check the status and output of a subtask by ID.',
+      parameters: {
+        type: 'object',
+        properties: {
+          task_id: { type: 'string', description: 'ID of the task to check' },
+        },
+        required: ['task_id'],
+      },
+    },
+  },
+
+  async execute(input: CheckTaskStatusInputT, _projectRoot: string, context: ToolContext): Promise<unknown> {
+    const { sql } = context;
+
+    const [task] = await sql<{ id: string; state: string; output_summary: string | null; started_at: string | null; ended_at: string | null }[]>`
+      SELECT id, state, output_summary, started_at, ended_at
+      FROM tasks
+      WHERE id = ${input.task_id}
+    `;
+
+    if (!task) {
+      return { error: `Task ${input.task_id} not found` };
+    }
+
+    return {
+      id: task.id,
+      state: task.state,
+      output_summary: task.output_summary,
+      started_at: task.started_at,
+      ended_at: task.ended_at,
+    };
+  },
+};
--- a/apps/coder/src/services/tools/create_file.ts
+++ b/apps/coder/src/services/tools/create_file.ts
@@ -0,0 +1,51 @@
+import { z } from 'zod';
+import type { ToolDef, ToolContext } from './types.js';
+import { queueCreate } from '../pending_changes.js';
+
+const CreateFileInput = z.object({
+  file_path: z.string().min(1),
+  content: z.string(),
+});
+type CreateFileInputT = z.infer<typeof CreateFileInput>;
+
+export const createFileTool: ToolDef<CreateFileInputT> = {
+  name: 'create_file',
+  description:
+    'Queue creation of a new file with the given content. ' +
+    'The change is staged in pending_changes and must be applied explicitly.',
+  inputSchema: CreateFileInput,
+  jsonSchema: {
+    type: 'function',
+    function: {
+      name: 'create_file',
+      description:
+        'Queue creation of a new file with the given content. ' +
+        'The change is staged in pending_changes and must be applied explicitly.',
+      parameters: {
+        type: 'object',
+        properties: {
+          file_path: { type: 'string', description: 'Path for the new file (relative to project root or absolute)' },
+          content: { type: 'string', description: 'Full content of the file to create' },
+        },
+        required: ['file_path', 'content'],
+      },
+    },
+  },
+  async execute(input: CreateFileInputT, projectRoot: string, context: ToolContext): Promise<unknown> {
+    const change = await queueCreate(
+      context.sql,
+      context.sessionId,
+      context.taskId,
+      input.file_path,
+      input.content,
+      projectRoot,
+    );
+    return {
+      status: 'queued',
+      change_id: change.id,
+      file_path: change.file_path,
+      operation: 'create',
+      message: `File creation queued: ${change.file_path}. Use apply_pending to write changes to disk.`,
+    };
+  },
+};
--- a/apps/coder/src/services/tools/delete_file.ts
+++ b/apps/coder/src/services/tools/delete_file.ts
@@ -0,0 +1,48 @@
+import { z } from 'zod';
+import type { ToolDef, ToolContext } from './types.js';
+import { queueDelete } from '../pending_changes.js';
+
+const DeleteFileInput = z.object({
+  file_path: z.string().min(1),
+});
+type DeleteFileInputT = z.infer<typeof DeleteFileInput>;
+
+export const deleteFileTool: ToolDef<DeleteFileInputT> = {
+  name: 'delete_file',
+  description:
+    'Queue deletion of a file. ' +
+    'The change is staged in pending_changes and must be applied explicitly.',
+  inputSchema: DeleteFileInput,
+  jsonSchema: {
+    type: 'function',
+    function: {
+      name: 'delete_file',
+      description:
+        'Queue deletion of a file. ' +
+        'The change is staged in pending_changes and must be applied explicitly.',
+      parameters: {
+        type: 'object',
+        properties: {
+          file_path: { type: 'string', description: 'Path to the file to delete (relative to project root or absolute)' },
+        },
+        required: ['file_path'],
+      },
+    },
+  },
+  async execute(input: DeleteFileInputT, projectRoot: string, context: ToolContext): Promise<unknown> {
+    const change = await queueDelete(
+      context.sql,
+      context.sessionId,
+      context.taskId,
+      input.file_path,
+      projectRoot,
+    );
+    return {
+      status: 'queued',
+      change_id: change.id,
+      file_path: change.file_path,
+      operation: 'delete',
+      message: `File deletion queued: ${change.file_path}. Use apply_pending to write changes to disk.`,
+    };
+  },
+};
--- a/apps/coder/src/services/tools/edit_file.ts
+++ b/apps/coder/src/services/tools/edit_file.ts
@@ -0,0 +1,54 @@
+import { z } from 'zod';
+import type { ToolDef, ToolContext } from './types.js';
+import { queueEdit } from '../pending_changes.js';
+
+const EditFileInput = z.object({
+  file_path: z.string().min(1),
+  old_string: z.string().min(1),
+  new_string: z.string(),
+});
+type EditFileInputT = z.infer<typeof EditFileInput>;
+
+export const editFileTool: ToolDef<EditFileInputT> = {
+  name: 'edit_file',
+  description:
+    'Queue an edit to a file. The edit replaces old_string with new_string. ' +
+    'The change is staged in pending_changes and must be applied explicitly.',
+  inputSchema: EditFileInput,
+  jsonSchema: {
+    type: 'function',
+    function: {
+      name: 'edit_file',
+      description:
+        'Queue an edit to a file. The edit replaces old_string with new_string. ' +
+        'The change is staged in pending_changes and must be applied explicitly.',
+      parameters: {
+        type: 'object',
+        properties: {
+          file_path: { type: 'string', description: 'Path to the file to edit (relative to project root or absolute)' },
+          old_string: { type: 'string', description: 'The exact string to find and replace (must appear in the file)' },
+          new_string: { type: 'string', description: 'The replacement string' },
+        },
+        required: ['file_path', 'old_string', 'new_string'],
+      },
+    },
+  },
+  async execute(input: EditFileInputT, projectRoot: string, context: ToolContext): Promise<unknown> {
+    const change = await queueEdit(
+      context.sql,
+      context.sessionId,
+      context.taskId,
+      input.file_path,
+      input.old_string,
+      input.new_string,
+      projectRoot,
+    );
+    return {
+      status: 'queued',
+      change_id: change.id,
+      file_path: change.file_path,
+      operation: 'edit',
+      message: `Edit queued for ${change.file_path}. Use apply_pending to write changes to disk.`,
+    };
+  },
+};
--- a/apps/coder/src/services/tools/index.ts
+++ b/apps/coder/src/services/tools/index.ts
@@ -0,0 +1,34 @@
+import type { ToolDef } from './types.js';
+import { editFileTool } from './edit_file.js';
+import { createFileTool } from './create_file.js';
+import { deleteFileTool } from './delete_file.js';
+import { applyPendingTool } from './apply_pending.js';
+import { rewindTool } from './rewind.js';
+import { newTaskTool } from './new_task.js';
+import { listTasksTool } from './list_tasks.js';
+import { checkTaskStatusTool } from './check_task_status.js';
+
+export type { ToolDef, ToolContext, ToolJsonSchema } from './types.js';
+
+// All BooCoder write tools. The inference loop (Phase 2B) will combine these
+// with BooChat's read-only tools to form the full tool set available to agents.
+// eslint-disable-next-line @typescript-eslint/no-explicit-any
+export const WRITE_TOOLS: readonly ToolDef<any>[] = [
+  applyPendingTool,
+  createFileTool,
+  deleteFileTool,
+  editFileTool,
+  rewindTool,
+  // Boomerang subtask tools — orchestrator agents call these to spawn/monitor child tasks.
+  // An "Orchestrator" agent profile would whitelist [new_task, list_tasks, check_task_status].
+  newTaskTool,
+  listTasksTool,
+  checkTaskStatusTool,
+];
+
+// eslint-disable-next-line @typescript-eslint/no-explicit-any
+export const WRITE_TOOLS_BY_NAME: ReadonlyMap<string, ToolDef<any>> = new Map(
+  WRITE_TOOLS.map((t) => [t.name, t]),
+);
+
+export { editFileTool, createFileTool, deleteFileTool, applyPendingTool, rewindTool, newTaskTool, listTasksTool, checkTaskStatusTool };
--- a/apps/coder/src/services/tools/inference_context.ts
+++ b/apps/coder/src/services/tools/inference_context.ts
@@ -0,0 +1,36 @@
+import type { Sql } from '../../db.js';
+
+/**
+ * Module-level inference context for write tools.
+ *
+ * Set via `setInferenceContext()` before each inference run starts.
+ * Write tools read it via `getInferenceContext()` during execute.
+ * Same pattern as BooChat's `loadConfig()` singleton — tools need
+ * ambient state that can't be threaded through the tool-phase execute
+ * signature (which is `execute(input, projectRoot, extraRoots?)`).
+ */
+
+export interface InferenceContext {
+  sql: Sql;
+  sessionId: string;
+  taskId: string | null;
+}
+
+let current: InferenceContext | null = null;
+
+export function setInferenceContext(ctx: InferenceContext): void {
+  current = ctx;
+}
+
+export function clearInferenceContext(): void {
+  current = null;
+}
+
+export function getInferenceContext(): InferenceContext {
+  if (!current) {
+    throw new Error(
+      'Write tool called outside inference context — setInferenceContext() was not called before this run',
+    );
+  }
+  return current;
+}
--- a/apps/coder/src/services/tools/list_tasks.ts
+++ b/apps/coder/src/services/tools/list_tasks.ts
@@ -0,0 +1,56 @@
+import { z } from 'zod';
+import type { ToolDef, ToolContext } from './types.js';
+import { getInferenceContext } from './inference_context.js';
+
+const ListTasksInput = z.object({
+  parent_task_id: z.string().uuid().optional().describe('Filter by parent task ID. Omit to list children of current task.'),
+});
+
+type ListTasksInputT = z.infer<typeof ListTasksInput>;
+
+export const listTasksTool: ToolDef<ListTasksInputT> = {
+  name: 'list_tasks',
+  description: 'List child tasks of the current task (or a specified parent). Returns id, state, input preview, and output_summary.',
+  inputSchema: ListTasksInput,
+  jsonSchema: {
+    type: 'function',
+    function: {
+      name: 'list_tasks',
+      description: 'List child tasks of the current task (or a specified parent).',
+      parameters: {
+        type: 'object',
+        properties: {
+          parent_task_id: { type: 'string', description: 'Filter by parent task ID. Omit to list children of current task.' },
+        },
+        required: [],
+      },
+    },
+  },
+
+  async execute(input: ListTasksInputT, _projectRoot: string, context: ToolContext): Promise<unknown> {
+    const { sql } = context;
+    const ctx = getInferenceContext();
+    const parentId = input.parent_task_id ?? ctx.taskId;
+
+    if (!parentId) {
+      return { tasks: [], note: 'No parent task context — not running inside a task.' };
+    }
+
+    const rows = await sql<{ id: string; state: string; input: string; output_summary: string | null }[]>`
+      SELECT id, state, input, output_summary
+      FROM tasks
+      WHERE parent_task_id = ${parentId}
+      ORDER BY created_at DESC
+      LIMIT 50
+    `;
+
+    return {
+      tasks: rows.map((r) => ({
+        id: r.id,
+        state: r.state,
+        input_preview: r.input.slice(0, 100),
+        output_summary: r.output_summary,
+      })),
+    };
+  },
+};
--- a/apps/coder/src/services/tools/new_task.ts
+++ b/apps/coder/src/services/tools/new_task.ts
@@ -0,0 +1,65 @@
+import { z } from 'zod';
+import type { ToolDef, ToolContext } from './types.js';
+import { getInferenceContext } from './inference_context.js';
+
+const NewTaskInput = z.object({
+  input: z.string().min(1).describe('Task description for the child subtask'),
+  agent: z.string().optional().describe('Optional: dispatch to a specific agent'),
+  model: z.string().optional().describe('Optional: model override for the subtask'),
+});
+
+type NewTaskInputT = z.infer<typeof NewTaskInput>;
+
+export const newTaskTool: ToolDef<NewTaskInputT> = {
+  name: 'new_task',
+  description:
+    'Spawn a subtask that runs in isolation. The subtask gets its own session and ' +
+    'worktree. Use check_task_status to monitor progress. Only the output_summary is ' +
+    'accessible to the parent — full isolation (Boomerang pattern).',
+  inputSchema: NewTaskInput,
+  jsonSchema: {
+    type: 'function',
+    function: {
+      name: 'new_task',
+      description:
+        'Spawn a subtask that runs in isolation. The subtask gets its own session and ' +
+        'worktree. Use check_task_status to monitor progress.',
+      parameters: {
+        type: 'object',
+        properties: {
+          input: { type: 'string', description: 'Task description for the child subtask' },
+          agent: { type: 'string', description: 'Optional: dispatch to a specific agent' },
+          model: { type: 'string', description: 'Optional: model override for the subtask' },
+        },
+        required: ['input'],
+      },
+    },
+  },
+
+  async execute(input: NewTaskInputT, _projectRoot: string, context: ToolContext): Promise<unknown> {
+    const { sql } = context;
+    // Get the current task's project_id from the inference context
+    const ctx = getInferenceContext();
+    const currentTaskId = ctx.taskId;
+
+    // Look up the project_id from the current session
+    const [session] = await sql<{ project_id: string }[]>`
+      SELECT project_id FROM sessions WHERE id = ${ctx.sessionId}
+    `;
+    if (!session) {
+      return { error: 'Cannot determine project_id from current session' };
+    }
+
+    const [task] = await sql<{ id: string; state: string }[]>`
+      INSERT INTO tasks (project_id, parent_task_id, input, agent, model)
+      VALUES (${session.project_id}, ${currentTaskId}, ${input.input}, ${input.agent ?? null}, ${input.model ?? null})
+      RETURNING id, state
+    `;
+
+    return {
+      message: `Subtask created (id: ${task!.id}). It will run in isolation. Use check_task_status to monitor.`,
+      task_id: task!.id,
+      state: task!.state,
+    };
+  },
+};
--- a/apps/coder/src/services/tools/rewind.ts
+++ b/apps/coder/src/services/tools/rewind.ts
@@ -0,0 +1,71 @@
+import { z } from 'zod';
+import type { ToolDef, ToolContext } from './types.js';
+import { rewindOne } from '../pending_changes.js';
+
+const RewindInput = z.object({
+  change_id: z.string().uuid().optional(),
+  all: z.boolean().optional(),
+});
+type RewindInputT = z.infer<typeof RewindInput>;
+
+export const rewindTool: ToolDef<RewindInputT> = {
+  name: 'rewind',
+  description:
+    'Revert applied changes. Provide change_id to revert a specific change, ' +
+    'or set all=true to revert all applied changes for the session (in reverse order).',
+  inputSchema: RewindInput,
+  jsonSchema: {
+    type: 'function',
+    function: {
+      name: 'rewind',
+      description:
+        'Revert applied changes. Provide change_id to revert a specific change, ' +
+        'or set all=true to revert all applied changes for the session (in reverse order).',
+      parameters: {
+        type: 'object',
+        properties: {
+          change_id: { type: 'string', format: 'uuid', description: 'ID of a specific change to revert' },
+          all: { type: 'boolean', description: 'If true, revert all applied changes for this session' },
+        },
+        required: [],
+      },
+    },
+  },
+  async execute(input: RewindInputT, projectRoot: string, context: ToolContext): Promise<unknown> {
+    if (input.change_id) {
+      const result = await rewindOne(context.sql, input.change_id, projectRoot);
+      return {
+        results: [result],
+        message: result.success
+          ? `Reverted change ${input.change_id} (${result.operation} on ${result.file_path}).`
+          : `Failed to revert: ${result.error}`,
+      };
+    }
+
+    if (input.all) {
+      // Rewind all applied changes for this session in reverse order
+      const applied = await context.sql<{ id: string }[]>`
+        SELECT id FROM pending_changes
+        WHERE session_id = ${context.sessionId} AND status = 'applied'
+        ORDER BY created_at DESC
+      `;
+      const results = [];
+      for (const row of applied) {
+        results.push(await rewindOne(context.sql, row.id, projectRoot));
+      }
+      const succeeded = results.filter((r) => r.success).length;
+      return {
+        total: results.length,
+        succeeded,
+        failed: results.length - succeeded,
+        results,
+        message:
+          results.length === 0
+            ? 'No applied changes to revert.'
+            : `Reverted ${succeeded}/${results.length} changes.`,
+      };
+    }
+
+    return { error: 'Provide either change_id or all=true.' };
+  },
+};
--- a/apps/coder/src/services/tools/types.ts
+++ b/apps/coder/src/services/tools/types.ts
@@ -0,0 +1,32 @@
+import type { z } from 'zod';
+import type { Sql } from '../../db.js';
+
+export interface ToolJsonSchema {
+  type: 'function';
+  function: {
+    name: string;
+    description: string;
+    parameters: Record<string, unknown>;
+  };
+}
+
+/**
+ * Context passed to BooCoder tool execute functions.
+ *
+ * Unlike BooChat's tools (which only need projectRoot), BooCoder's write tools
+ * interact with the database (pending_changes table) and need session/task
+ * context for proper attribution.
+ */
+export interface ToolContext {
+  sql: Sql;
+  sessionId: string;
+  taskId: string | null;
+}
+
+export interface ToolDef<TInput> {
+  name: string;
+  description: string;
+  inputSchema: z.ZodType<TInput>;
+  jsonSchema: ToolJsonSchema;
+  execute(input: TInput, projectRoot: string, context: ToolContext): Promise<unknown>;
+}
--- a/apps/coder/src/services/worktrees.ts
+++ b/apps/coder/src/services/worktrees.ts
@@ -0,0 +1,118 @@
+/**
+ * Git worktree management for external agent dispatch.
+ *
+ * Each dispatched task gets its own git worktree so the external agent
+ * can modify files freely without touching the main working tree.
+ * After the agent completes, we diff the worktree against HEAD and
+ * queue the diff into pending_changes.
+ */
+import { sshExec } from './ssh.js';
+
+const WORKTREE_BASE = '/tmp/booworktrees';
+
+/**
+ * Create a git worktree for a task on the host.
+ * Returns the absolute path to the worktree directory.
+ */
+export async function createWorktree(
+  projectPath: string,
+  taskId: string,
+  opts?: { signal?: AbortSignal },
+): Promise<string> {
+  const worktreePath = `${WORKTREE_BASE}/${taskId}`;
+  const branchName = `task-${taskId}`;
+
+  // Ensure the base directory exists
+  await sshExec(`mkdir -p ${WORKTREE_BASE}`, { signal: opts?.signal });
+
+  // Create the worktree with a new branch from HEAD
+  const result = await sshExec(
+    `git -C ${shellEscape(projectPath)} worktree add ${shellEscape(worktreePath)} -b ${shellEscape(branchName)} HEAD`,
+    { signal: opts?.signal, timeoutMs: 30_000 },
+  );
+
+  if (result.exitCode !== 0) {
+    throw new Error(`Failed to create worktree: ${result.stderr.trim() || result.stdout.trim()}`);
+  }
+
+  return worktreePath;
+}
+
+/**
+ * Get the unified diff of changes made in the worktree vs the parent branch (HEAD).
+ * Returns an empty string if there are no changes.
+ */
+export async function diffWorktree(
+  worktreePath: string,
+  projectPath: string,
+  opts?: { signal?: AbortSignal },
+): Promise<string> {
+  // First, commit any uncommitted changes in the worktree so we can diff branches
+  // Stage all changes
+  const addResult = await sshExec(
+    `cd ${shellEscape(worktreePath)} && git add -A`,
+    { signal: opts?.signal, timeoutMs: 30_000 },
+  );
+  if (addResult.exitCode !== 0) {
+    throw new Error(`Failed to stage worktree changes: ${addResult.stderr.trim()}`);
+  }
+
+  // Check if there are staged changes
+  const statusResult = await sshExec(
+    `cd ${shellEscape(worktreePath)} && git diff --cached --quiet`,
+    { signal: opts?.signal, timeoutMs: 10_000 },
+  );
+
+  if (statusResult.exitCode === 0) {
+    // No changes
+    return '';
+  }
+
+  // Commit staged changes (needed to produce a clean branch diff)
+  await sshExec(
+    `cd ${shellEscape(worktreePath)} && git -c user.email=boocoder@local -c user.name=BooCoder commit -m "task changes" --allow-empty`,
+    { signal: opts?.signal, timeoutMs: 15_000 },
+  );
+
+  // Diff the worktree branch against the parent commit (HEAD of main tree)
+  const diffResult = await sshExec(
+    `git -C ${shellEscape(projectPath)} diff HEAD...$(git -C ${shellEscape(worktreePath)} rev-parse HEAD)`,
+    { signal: opts?.signal, timeoutMs: 60_000 },
+  );
+
+  if (diffResult.exitCode !== 0) {
+    throw new Error(`Failed to diff worktree: ${diffResult.stderr.trim()}`);
+  }
+
+  return diffResult.stdout;
+}
+
+/**
+ * Remove a worktree and its associated branch.
+ * Best-effort — does not throw on failure (task may have already been cleaned up).
+ */
+export async function cleanupWorktree(
+  projectPath: string,
+  taskId: string,
+): Promise<void> {
+  const worktreePath = `${WORKTREE_BASE}/${taskId}`;
+  const branchName = `task-${taskId}`;
+
+  // Remove the worktree (--force handles dirty state)
+  await sshExec(
+    `git -C ${shellEscape(projectPath)} worktree remove ${shellEscape(worktreePath)} --force`,
+    { timeoutMs: 15_000 },
+  ).catch(() => {});
+
+  // Delete the task branch
+  await sshExec(
+    `git -C ${shellEscape(projectPath)} branch -D ${shellEscape(branchName)}`,
+    { timeoutMs: 10_000 },
+  ).catch(() => {});
+}
+
+/** Minimal shell escape for paths (single-quote wrapping). */
+function shellEscape(s: string): string {
+  // Replace single quotes with escaped version, wrap in single quotes
+  return "'" + s.replace(/'/g, "'\\''") + "'";
+}
--- a/apps/coder/src/services/write_guard.ts
+++ b/apps/coder/src/services/write_guard.ts
@@ -0,0 +1,77 @@
+import { resolve, sep } from 'node:path';
+
+export class WriteGuardError extends Error {
+  constructor(message: string) {
+    super(message);
+    this.name = 'WriteGuardError';
+  }
+}
+
+// Deny list: files that should never be written regardless of path-guard.
+// Subset of BooChat's secret_guard.ts — covers the most dangerous patterns.
+// Full parity with BooChat's deny list is not needed for write-guard because
+// the write tools are intentional (model chose to create/edit); we block only
+// files that are unambiguously secrets.
+const SECRET_PATTERNS: readonly string[] = [
+  '.env',
+  '.env.local',
+  '.env.production',
+  '.env.development',
+  '.env.staging',
+  'id_rsa',
+  'id_dsa',
+  'id_ecdsa',
+  'id_ed25519',
+  '*.pem',
+  '*.key',
+  '*.p12',
+  '*.pfx',
+  '*.crt',
+  'credentials.json',
+  '*.kdbx',
+  '.netrc',
+];
+
+export function isSecretPath(filePath: string): boolean {
+  const normalized = filePath.replace(/\\/g, '/');
+  const segments = normalized.split('/').filter((s) => s.length > 0);
+  if (segments.length === 0) return false;
+  const basename = segments[segments.length - 1]!;
+
+  return SECRET_PATTERNS.some((pattern) => {
+    if (pattern.startsWith('*')) {
+      return basename.endsWith(pattern.slice(1));
+    }
+    return basename === pattern;
+  });
+}
+
+/**
+ * Resolve and validate a write target path.
+ *
+ * Key difference from BooChat's pathGuard: no realpath() — the file may not
+ * exist yet (creates). Uses resolve() to normalize ../ segments and then
+ * checks the result stays within projectRoot.
+ */
+export function resolveWritePath(projectRoot: string, filePath: string): string {
+  if (!filePath || filePath.trim().length === 0) {
+    throw new WriteGuardError('file path is required');
+  }
+
+  if (filePath.includes('\x00')) {
+    throw new WriteGuardError('file path contains null byte');
+  }
+
+  const candidate = filePath.startsWith('/') ? filePath : resolve(projectRoot, filePath);
+  const normalized = resolve(candidate); // normalizes ../ segments
+
+  if (!normalized.startsWith(projectRoot + sep) && normalized !== projectRoot) {
+    throw new WriteGuardError(`path escapes project root: ${filePath}`);
+  }
+
+  if (isSecretPath(normalized)) {
+    throw new WriteGuardError(`cannot write to secret file: ${filePath}`);
+  }
+
+  return normalized;
+}
--- a/apps/coder/tsconfig.json
+++ b/apps/coder/tsconfig.json
@@ -0,0 +1,15 @@
+{
+  "extends": "../../tsconfig.base.json",
+  "compilerOptions": {
+    "module": "NodeNext",
+    "moduleResolution": "NodeNext",
+    "outDir": "dist",
+    "rootDir": "src",
+    "lib": ["ES2022"],
+    "types": ["node"],
+    "declaration": false,
+    "sourceMap": true
+  },
+  "include": ["src/**/*"],
+  "exclude": ["src/**/__tests__/**", "**/*.test.ts"]
+}
--- a/apps/coder/vitest.config.ts
+++ b/apps/coder/vitest.config.ts
@@ -0,0 +1,9 @@
+import { defineConfig } from 'vitest/config';
+
+export default defineConfig({
+  test: {
+    environment: 'node',
+    globals: false,
+    include: ['src/**/__tests__/**/*.test.ts'],
+  },
+});
--- a/apps/coder/web/index.html
+++ b/apps/coder/web/index.html
@@ -0,0 +1,12 @@
+<!doctype html>
+<html lang="en" class="dark">
+  <head>
+    <meta charset="UTF-8" />
+    <meta name="viewport" content="width=device-width, initial-scale=1.0" />
+    <title>BooCoder</title>
+  </head>
+  <body class="bg-zinc-900 text-zinc-100">
+    <div id="root"></div>
+    <script type="module" src="/src/main.tsx"></script>
+  </body>
+</html>
--- a/apps/coder/web/package.json
+++ b/apps/coder/web/package.json
@@ -0,0 +1,29 @@
+{
+  "name": "@boocode/coder-web",
+  "version": "2.0.0",
+  "private": true,
+  "type": "module",
+  "scripts": {
+    "dev": "vite",
+    "build": "tsc -b && vite build",
+    "typecheck": "tsc -b --noEmit",
+    "preview": "vite preview"
+  },
+  "dependencies": {
+    "lucide-react": "^1.16.0",
+    "react": "^18.3.1",
+    "react-dom": "^18.3.1",
+    "react-markdown": "^10.1.0",
+    "react-router-dom": "^6.26.0",
+    "remark-gfm": "^4.0.1"
+  },
+  "devDependencies": {
+    "@tailwindcss/postcss": "^4.3.0",
+    "@types/react": "^18.3.3",
+    "@types/react-dom": "^18.3.0",
+    "@vitejs/plugin-react": "^4.3.1",
+    "tailwindcss": "^4.3.0",
+    "typescript": "^5.5.0",
+    "vite": "^5.3.4"
+  }
+}
--- a/apps/coder/web/postcss.config.js
+++ b/apps/coder/web/postcss.config.js
@@ -0,0 +1,5 @@
+export default {
+  plugins: {
+    '@tailwindcss/postcss': {},
+  },
+};
--- a/apps/coder/web/src/App.tsx
+++ b/apps/coder/web/src/App.tsx
@@ -0,0 +1,13 @@
+import { Routes, Route, Navigate } from 'react-router-dom';
+import { Home } from './pages/Home';
+import { Session } from './pages/Session';
+
+export function App() {
+  return (
+    <Routes>
+      <Route path="/" element={<Home />} />
+      <Route path="/sessions/:sessionId" element={<Session />} />
+      <Route path="*" element={<Navigate to="/" replace />} />
+    </Routes>
+  );
+}
--- a/apps/coder/web/src/api/client.ts
+++ b/apps/coder/web/src/api/client.ts
@@ -0,0 +1,93 @@
+import type { Project, Session, Chat, Message, PendingChange } from './types';
+
+export class ApiError extends Error {
+  constructor(
+    public status: number,
+    public body: unknown,
+  ) {
+    super(
+      typeof body === 'object' && body && 'error' in body
+        ? String((body as { error: unknown }).error)
+        : `HTTP ${status}`,
+    );
+  }
+}
+
+async function request<T>(path: string, init: RequestInit = {}): Promise<T> {
+  const res = await fetch(path, {
+    ...init,
+    headers: {
+      'Content-Type': 'application/json',
+      ...(init.headers ?? {}),
+    },
+  });
+  if (res.status === 204) return undefined as T;
+  const text = await res.text();
+  const data = text ? JSON.parse(text) : undefined;
+  if (!res.ok) throw new ApiError(res.status, data);
+  return data as T;
+}
+
+export const api = {
+  health: () => request<{ ok: boolean; db: boolean; tools: number }>('/api/health'),
+
+  projects: {
+    list: (params?: { status?: 'open' | 'archived' }) =>
+      request<Project[]>(`/api/projects${params?.status ? `?status=${params.status}` : ''}`),
+  },
+
+  sessions: {
+    listForProject: (projectId: string, status?: 'open' | 'archived') =>
+      request<Session[]>(
+        `/api/projects/${projectId}/sessions${status ? `?status=${status}` : ''}`,
+      ),
+    get: (id: string) => request<Session>(`/api/sessions/${id}`),
+  },
+
+  chats: {
+    listForSession: (sessionId: string) =>
+      request<Chat[]>(`/api/sessions/${sessionId}/chats`),
+    create: (sessionId: string, body?: { name?: string }) =>
+      request<Chat>(`/api/sessions/${sessionId}/chats`, {
+        method: 'POST',
+        body: JSON.stringify(body ?? {}),
+      }),
+  },
+
+  messages: {
+    send: (sessionId: string, chatId: string, content: string) =>
+      request<{ user_message_id: string; assistant_message_id: string }>(
+        `/api/sessions/${sessionId}/messages`,
+        {
+          method: 'POST',
+          body: JSON.stringify({ content, chat_id: chatId }),
+        },
+      ),
+    stop: (sessionId: string) =>
+      request<{ cancelled: boolean }>(`/api/sessions/${sessionId}/stop`, {
+        method: 'POST',
+      }),
+  },
+
+  pending: {
+    list: (sessionId: string) =>
+      request<PendingChange[]>(`/api/sessions/${sessionId}/pending`),
+    applyAll: (sessionId: string) =>
+      request<{ results: Array<{ id: string; success: boolean; error?: string }> }>(
+        `/api/sessions/${sessionId}/pending/apply`,
+        { method: 'POST' },
+      ),
+    applyOne: (changeId: string) =>
+      request<{ success: boolean; error?: string }>(`/api/pending/${changeId}/apply`, {
+        method: 'POST',
+      }),
+    rejectOne: (changeId: string) =>
+      request<{ ok: boolean }>(`/api/pending/${changeId}/reject`, {
+        method: 'POST',
+      }),
+    rewindOne: (changeId: string) =>
+      request<{ success: boolean; error?: string }>(`/api/pending/${changeId}/rewind`, {
+        method: 'POST',
+      }),
+  },
+};
--- a/apps/coder/web/src/api/types.ts
+++ b/apps/coder/web/src/api/types.ts
@@ -0,0 +1,89 @@
+// Minimal types for the BooCoder frontend.
+// Shared DB entities (same schema as BooChat).
+
+export interface Project {
+  id: string;
+  name: string;
+  path: string;
+  status: 'open' | 'archived';
+  created_at: string;
+  updated_at: string;
+}
+
+export interface Session {
+  id: string;
+  project_id: string;
+  name: string | null;
+  model: string | null;
+  status: 'open' | 'archived';
+  created_at: string;
+  updated_at: string;
+}
+
+export interface Chat {
+  id: string;
+  session_id: string;
+  name: string | null;
+  status: 'open' | 'archived';
+  created_at: string;
+  updated_at: string;
+}
+
+export interface ToolCall {
+  id: string;
+  name: string;
+  arguments: string;
+}
+
+export interface ToolResult {
+  tool_call_id: string;
+  output: string;
+  truncated?: boolean;
+  error?: boolean;
+}
+
+export interface Message {
+  id: string;
+  session_id: string;
+  chat_id: string;
+  role: 'user' | 'assistant' | 'tool' | 'system';
+  content: string;
+  kind: string;
+  tool_calls: ToolCall[] | null;
+  tool_results: ToolResult | null;
+  status: 'streaming' | 'complete' | 'failed' | 'cancelled';
+  tokens_used: number | null;
+  ctx_used: number | null;
+  ctx_max: number | null;
+  started_at: string | null;
+  finished_at: string | null;
+  created_at: string;
+  metadata: unknown;
+}
+
+export interface PendingChange {
+  id: string;
+  session_id: string;
+  task_id: string | null;
+  file_path: string;
+  operation: 'create' | 'edit' | 'delete';
+  old_string: string | null;
+  new_string: string | null;
+  content: string | null;
+  diff: string | null;
+  status: 'pending' | 'applied' | 'rejected' | 'reverted';
+  created_at: string;
+  applied_at: string | null;
+}
+
+// WebSocket frame types (subset of what the coder backend publishes)
+export type WsFrame =
+  | { type: 'snapshot'; messages: Message[] }
+  | { type: 'message_started'; message_id: string; chat_id: string; role: Message['role'] }
+  | { type: 'delta'; message_id: string; chat_id: string; content: string }
+  | { type: 'tool_call'; message_id: string; chat_id: string; tool_call: ToolCall }
+  | { type: 'tool_result'; tool_message_id: string; chat_id: string; tool_call_id: string; output: string; truncated?: boolean; error?: boolean }
+  | { type: 'message_complete'; message_id: string; chat_id: string; tokens_used?: number; ctx_used?: number; ctx_max?: number; started_at?: string; finished_at?: string; metadata?: unknown }
+  | { type: 'error'; message_id?: string; error: string; reason?: string }
+  | { type: 'pending_change_added'; change: PendingChange }
+  | { type: 'pending_change_updated'; change: PendingChange };
--- a/apps/coder/web/src/components/ChatPane.tsx
+++ b/apps/coder/web/src/components/ChatPane.tsx
@@ -0,0 +1,131 @@
+import { useState, useRef, useEffect } from 'react';
+import { Send, Square } from 'lucide-react';
+import type { Message } from '@/api/types';
+import { api } from '@/api/client';
+import { MessageBubble } from './MessageBubble';
+
+interface Props {
+  sessionId: string;
+  chatId: string;
+  messages: Message[];
+  isStreaming: boolean;
+  connected: boolean;
+}
+
+export function ChatPane({ sessionId, chatId, messages, isStreaming, connected }: Props) {
+  const [input, setInput] = useState('');
+  const [sending, setSending] = useState(false);
+  const messagesEndRef = useRef<HTMLDivElement>(null);
+  const textareaRef = useRef<HTMLTextAreaElement>(null);
+
+  // Auto-scroll to bottom when messages change
+  useEffect(() => {
+    messagesEndRef.current?.scrollIntoView({ behavior: 'smooth' });
+  }, [messages]);
+
+  // Auto-resize textarea
+  useEffect(() => {
+    const el = textareaRef.current;
+    if (!el) return;
+    el.style.height = 'auto';
+    el.style.height = Math.min(el.scrollHeight, 200) + 'px';
+  }, [input]);
+
+  const handleSend = async () => {
+    const content = input.trim();
+    if (!content || sending || isStreaming) return;
+
+    setInput('');
+    setSending(true);
+    try {
+      await api.messages.send(sessionId, chatId, content);
+    } catch (err) {
+      console.error('send failed:', err);
+      // Restore input on failure
+      setInput(content);
+    } finally {
+      setSending(false);
+    }
+  };
+
+  const handleStop = async () => {
+    try {
+      await api.messages.stop(sessionId);
+    } catch (err) {
+      console.error('stop failed:', err);
+    }
+  };
+
+  const handleKeyDown = (e: React.KeyboardEvent) => {
+    if (e.key === 'Enter' && !e.shiftKey) {
+      e.preventDefault();
+      handleSend();
+    }
+  };
+
+  // Filter out system messages for display (sentinels)
+  const visibleMessages = messages.filter((m) => m.role !== 'system');
+
+  return (
+    <div className="flex flex-col h-full">
+      {/* Connection indicator */}
+      <div className="flex items-center gap-2 px-4 py-2 border-b border-zinc-800 text-xs text-zinc-500">
+        <div
+          className={`w-1.5 h-1.5 rounded-full ${connected ? 'bg-green-500' : 'bg-red-500'}`}
+        />
+        <span>{connected ? 'Connected' : 'Disconnected'}</span>
+        {isStreaming && (
+          <span className="text-blue-400 ml-auto">Generating...</span>
+        )}
+      </div>
+
+      {/* Messages list */}
+      <div className="flex-1 overflow-y-auto px-4 py-4">
+        {visibleMessages.length === 0 && (
+          <div className="text-center text-zinc-500 mt-8">
+            <p className="text-lg font-medium">BooCoder</p>
+            <p className="text-sm mt-1">Send a message to start coding.</p>
+          </div>
+        )}
+        {visibleMessages.map((msg) => (
+          <MessageBubble key={msg.id} message={msg} />
+        ))}
+        <div ref={messagesEndRef} />
+      </div>
+
+      {/* Input area */}
+      <div className="border-t border-zinc-800 px-4 py-3">
+        <div className="flex items-end gap-2">
+          <textarea
+            ref={textareaRef}
+            value={input}
+            onChange={(e) => setInput(e.target.value)}
+            onKeyDown={handleKeyDown}
+            placeholder="Message BooCoder..."
+            rows={1}
+            className="flex-1 bg-zinc-800 border border-zinc-700 rounded-lg px-3 py-2 text-sm text-zinc-100 placeholder-zinc-500 resize-none focus:outline-none focus:ring-1 focus:ring-blue-500 focus:border-blue-500"
+            disabled={sending}
+          />
+          {isStreaming ? (
+            <button
+              onClick={handleStop}
+              className="p-2 rounded-lg bg-red-600 hover:bg-red-500 text-white transition-colors"
+              title="Stop generation"
+            >
+              <Square size={18} />
+            </button>
+          ) : (
+            <button
+              onClick={handleSend}
+              disabled={!input.trim() || sending}
+              className="p-2 rounded-lg bg-blue-600 hover:bg-blue-500 disabled:opacity-40 disabled:cursor-not-allowed text-white transition-colors"
+              title="Send message"
+            >
+              <Send size={18} />
+            </button>
+          )}
+        </div>
+      </div>
+    </div>
+  );
+}
--- a/apps/coder/web/src/components/DiffPane.tsx
+++ b/apps/coder/web/src/components/DiffPane.tsx
@@ -0,0 +1,352 @@
+import { useState, useEffect, useCallback } from 'react';
+import { Check, X, RotateCcw, FileText, FilePlus, Trash2, RefreshCw } from 'lucide-react';
+import type { PendingChange } from '@/api/types';
+import { api } from '@/api/client';
+
+interface Props {
+  sessionId: string;
+  onPendingChange: (cb: (change: PendingChange) => void) => () => void;
+}
+
+export function DiffPane({ sessionId, onPendingChange }: Props) {
+  const [changes, setChanges] = useState<PendingChange[]>([]);
+  const [loading, setLoading] = useState(true);
+  const [expandedId, setExpandedId] = useState<string | null>(null);
+
+  const fetchPending = useCallback(async () => {
+    try {
+      const result = await api.pending.list(sessionId);
+      setChanges(result);
+    } catch (err) {
+      console.error('fetch pending failed:', err);
+    } finally {
+      setLoading(false);
+    }
+  }, [sessionId]);
+
+  // Initial load
+  useEffect(() => {
+    fetchPending();
+  }, [fetchPending]);
+
+  // Listen for WS pending change events
+  useEffect(() => {
+    const unsub = onPendingChange((change) => {
+      setChanges((prev) => {
+        const idx = prev.findIndex((c) => c.id === change.id);
+        if (idx >= 0) {
+          const next = [...prev];
+          next[idx] = change;
+          return next;
+        }
+        return [...prev, change];
+      });
+    });
+    return unsub;
+  }, [onPendingChange]);
+
+  const pendingChanges = changes.filter((c) => c.status === 'pending');
+  const resolvedChanges = changes.filter((c) => c.status !== 'pending');
+
+  const handleApplyOne = async (id: string) => {
+    try {
+      await api.pending.applyOne(id);
+      setChanges((prev) =>
+        prev.map((c) => (c.id === id ? { ...c, status: 'applied' as const } : c)),
+      );
+    } catch (err) {
+      console.error('apply failed:', err);
+    }
+  };
+
+  const handleRejectOne = async (id: string) => {
+    try {
+      await api.pending.rejectOne(id);
+      setChanges((prev) =>
+        prev.map((c) => (c.id === id ? { ...c, status: 'rejected' as const } : c)),
+      );
+    } catch (err) {
+      console.error('reject failed:', err);
+    }
+  };
+
+  const handleRewindOne = async (id: string) => {
+    try {
+      await api.pending.rewindOne(id);
+      setChanges((prev) =>
+        prev.map((c) => (c.id === id ? { ...c, status: 'reverted' as const } : c)),
+      );
+    } catch (err) {
+      console.error('rewind failed:', err);
+    }
+  };
+
+  const handleApplyAll = async () => {
+    try {
+      const result = await api.pending.applyAll(sessionId);
+      const appliedIds = new Set(
+        result.results.filter((r) => r.success).map((r) => r.id),
+      );
+      setChanges((prev) =>
+        prev.map((c) =>
+          appliedIds.has(c.id) ? { ...c, status: 'applied' as const } : c,
+        ),
+      );
+    } catch (err) {
+      console.error('apply all failed:', err);
+    }
+  };
+
+  const handleRejectAll = async () => {
+    // Reject each pending change individually (no batch reject endpoint)
+    for (const c of pendingChanges) {
+      await handleRejectOne(c.id);
+    }
+  };
+
+  const OpIcon = ({ op }: { op: PendingChange['operation'] }) => {
+    switch (op) {
+      case 'create':
+        return <FilePlus size={14} className="text-green-400" />;
+      case 'edit':
+        return <FileText size={14} className="text-blue-400" />;
+      case 'delete':
+        return <Trash2 size={14} className="text-red-400" />;
+    }
+  };
+
+  const StatusBadge = ({ status }: { status: PendingChange['status'] }) => {
+    const colors: Record<PendingChange['status'], string> = {
+      pending: 'bg-yellow-500/20 text-yellow-400',
+      applied: 'bg-green-500/20 text-green-400',
+      rejected: 'bg-zinc-500/20 text-zinc-400',
+      reverted: 'bg-orange-500/20 text-orange-400',
+    };
+    return (
+      <span className={`text-[10px] px-1.5 py-0.5 rounded ${colors[status]}`}>
+        {status}
+      </span>
+    );
+  };
+
+  return (
+    <div className="flex flex-col h-full">
+      {/* Header */}
+      <div className="flex items-center justify-between px-4 py-2 border-b border-zinc-800">
+        <h2 className="text-sm font-medium text-zinc-300">
+          Pending Changes
+          {pendingChanges.length > 0 && (
+            <span className="ml-1.5 text-xs text-zinc-500">
+              ({pendingChanges.length})
+            </span>
+          )}
+        </h2>
+        <div className="flex items-center gap-1">
+          <button
+            onClick={fetchPending}
+            className="p-1 rounded hover:bg-zinc-800 text-zinc-400 hover:text-zinc-200"
+            title="Refresh"
+          >
+            <RefreshCw size={14} />
+          </button>
+          {pendingChanges.length > 0 && (
+            <>
+              <button
+                onClick={handleApplyAll}
+                className="text-xs px-2 py-1 rounded bg-green-600/80 hover:bg-green-600 text-white"
+              >
+                Apply All
+              </button>
+              <button
+                onClick={handleRejectAll}
+                className="text-xs px-2 py-1 rounded bg-zinc-700 hover:bg-zinc-600 text-zinc-300"
+              >
+                Reject All
+              </button>
+            </>
+          )}
+        </div>
+      </div>
+
+      {/* Changes list */}
+      <div className="flex-1 overflow-y-auto">
+        {loading && (
+          <div className="text-center text-zinc-500 text-sm py-8">Loading...</div>
+        )}
+
+        {!loading && changes.length === 0 && (
+          <div className="text-center text-zinc-500 text-sm py-8">
+            No pending changes yet.
+          </div>
+        )}
+
+        {/* Pending changes first */}
+        {pendingChanges.map((change) => (
+          <ChangeItem
+            key={change.id}
+            change={change}
+            expanded={expandedId === change.id}
+            onToggle={() =>
+              setExpandedId((prev) => (prev === change.id ? null : change.id))
+            }
+            onApply={() => handleApplyOne(change.id)}
+            onReject={() => handleRejectOne(change.id)}
+            OpIcon={OpIcon}
+            StatusBadge={StatusBadge}
+          />
+        ))}
+
+        {/* Resolved changes */}
+        {resolvedChanges.length > 0 && pendingChanges.length > 0 && (
+          <div className="border-t border-zinc-800 my-1" />
+        )}
+        {resolvedChanges.map((change) => (
+          <ChangeItem
+            key={change.id}
+            change={change}
+            expanded={expandedId === change.id}
+            onToggle={() =>
+              setExpandedId((prev) => (prev === change.id ? null : change.id))
+            }
+            onRewind={
+              change.status === 'applied'
+                ? () => handleRewindOne(change.id)
+                : undefined
+            }
+            OpIcon={OpIcon}
+            StatusBadge={StatusBadge}
+          />
+        ))}
+      </div>
+    </div>
+  );
+}
+
+interface ChangeItemProps {
+  change: PendingChange;
+  expanded: boolean;
+  onToggle: () => void;
+  onApply?: () => void;
+  onReject?: () => void;
+  onRewind?: () => void;
+  OpIcon: React.ComponentType<{ op: PendingChange['operation'] }>;
+  StatusBadge: React.ComponentType<{ status: PendingChange['status'] }>;
+}
+
+function ChangeItem({
+  change,
+  expanded,
+  onToggle,
+  onApply,
+  onReject,
+  onRewind,
+  OpIcon,
+  StatusBadge,
+}: ChangeItemProps) {
+  const fileName = change.file_path.split('/').pop() || change.file_path;
+  const dirPath = change.file_path.split('/').slice(0, -1).join('/');
+
+  return (
+    <div className="border-b border-zinc-800/50">
+      <div
+        className="flex items-center gap-2 px-4 py-2 hover:bg-zinc-800/50 cursor-pointer"
+        onClick={onToggle}
+      >
+        <OpIcon op={change.operation} />
+        <div className="flex-1 min-w-0">
+          <span className="text-sm font-mono text-zinc-200 truncate block">
+            {fileName}
+          </span>
+          {dirPath && (
+            <span className="text-[11px] text-zinc-500 truncate block">
+              {dirPath}
+            </span>
+          )}
+        </div>
+        <StatusBadge status={change.status} />
+        {change.status === 'pending' && (
+          <div className="flex items-center gap-1 ml-1">
+            <button
+              onClick={(e) => {
+                e.stopPropagation();
+                onApply?.();
+              }}
+              className="p-1 rounded hover:bg-green-600/30 text-green-400"
+              title="Apply"
+            >
+              <Check size={14} />
+            </button>
+            <button
+              onClick={(e) => {
+                e.stopPropagation();
+                onReject?.();
+              }}
+              className="p-1 rounded hover:bg-red-600/30 text-red-400"
+              title="Reject"
+            >
+              <X size={14} />
+            </button>
+          </div>
+        )}
+        {change.status === 'applied' && onRewind && (
+          <button
+            onClick={(e) => {
+              e.stopPropagation();
+              onRewind();
+            }}
+            className="p-1 rounded hover:bg-orange-600/30 text-orange-400"
+            title="Rewind"
+          >
+            <RotateCcw size={14} />
+          </button>
+        )}
+      </div>
+
+      {expanded && (
+        <div className="px-4 pb-3">
+          {change.operation === 'edit' && (
+            <div className="space-y-2">
+              {change.old_string && (
+                <div className="rounded bg-red-950/30 border border-red-900/30 p-2">
+                  <div className="text-[10px] text-red-400 mb-1 font-medium">
+                    Remove
+                  </div>
+                  <pre className="text-xs text-red-200 whitespace-pre-wrap break-all font-mono">
+                    {change.old_string}
+                  </pre>
+                </div>
+              )}
+              {change.new_string && (
+                <div className="rounded bg-green-950/30 border border-green-900/30 p-2">
+                  <div className="text-[10px] text-green-400 mb-1 font-medium">
+                    Add
+                  </div>
+                  <pre className="text-xs text-green-200 whitespace-pre-wrap break-all font-mono">
+                    {change.new_string}
+                  </pre>
+                </div>
+              )}
+            </div>
+          )}
+          {change.operation === 'create' && change.content && (
+            <div className="rounded bg-green-950/30 border border-green-900/30 p-2">
+              <div className="text-[10px] text-green-400 mb-1 font-medium">
+                New file
+              </div>
+              <pre className="text-xs text-green-200 whitespace-pre-wrap break-all font-mono max-h-60 overflow-y-auto">
+                {change.content.length > 2000
+                  ? change.content.slice(0, 2000) + '\n... (truncated)'
+                  : change.content}
+              </pre>
+            </div>
+          )}
+          {change.operation === 'delete' && (
+            <div className="rounded bg-red-950/30 border border-red-900/30 p-2 text-xs text-red-300">
+              This file will be deleted.
+            </div>
+          )}
+        </div>
+      )}
+    </div>
+  );
+}
--- a/apps/coder/web/src/components/Layout.tsx
+++ b/apps/coder/web/src/components/Layout.tsx
@@ -0,0 +1,62 @@
+import { useState } from 'react';
+import { Code2, MessageSquare, GitPullRequest } from 'lucide-react';
+
+interface Props {
+  chatPane: React.ReactNode;
+  diffPane: React.ReactNode;
+}
+
+export function Layout({ chatPane, diffPane }: Props) {
+  const [activeTab, setActiveTab] = useState<'chat' | 'diff'>('chat');
+
+  return (
+    <div className="flex flex-col h-screen bg-zinc-900">
+      {/* Top bar */}
+      <header className="flex items-center gap-3 px-4 py-2 border-b border-zinc-800 bg-zinc-900/95">
+        <Code2 size={20} className="text-blue-400" />
+        <h1 className="text-sm font-semibold text-zinc-200">BooCoder</h1>
+      </header>
+
+      {/* Mobile tab bar (visible below lg breakpoint) */}
+      <div className="lg:hidden flex border-b border-zinc-800">
+        <button
+          onClick={() => setActiveTab('chat')}
+          className={`flex-1 flex items-center justify-center gap-1.5 py-2 text-sm ${
+            activeTab === 'chat'
+              ? 'text-blue-400 border-b-2 border-blue-400'
+              : 'text-zinc-500'
+          }`}
+        >
+          <MessageSquare size={14} />
+          Chat
+        </button>
+        <button
+          onClick={() => setActiveTab('diff')}
+          className={`flex-1 flex items-center justify-center gap-1.5 py-2 text-sm ${
+            activeTab === 'diff'
+              ? 'text-blue-400 border-b-2 border-blue-400'
+              : 'text-zinc-500'
+          }`}
+        >
+          <GitPullRequest size={14} />
+          Changes
+        </button>
+      </div>
+
+      {/* Desktop split layout */}
+      <div className="flex-1 hidden lg:flex overflow-hidden">
+        <div className="w-[60%] border-r border-zinc-800 overflow-hidden">
+          {chatPane}
+        </div>
+        <div className="w-[40%] overflow-hidden">
+          {diffPane}
+        </div>
+      </div>
+
+      {/* Mobile: show only the active tab */}
+      <div className="flex-1 lg:hidden overflow-hidden">
+        {activeTab === 'chat' ? chatPane : diffPane}
+      </div>
+    </div>
+  );
+}
--- a/apps/coder/web/src/components/MessageBubble.tsx
+++ b/apps/coder/web/src/components/MessageBubble.tsx
@@ -0,0 +1,115 @@
+import Markdown from 'react-markdown';
+import remarkGfm from 'remark-gfm';
+import type { Message } from '@/api/types';
+import { Wrench, AlertCircle, Loader2 } from 'lucide-react';
+
+interface Props {
+  message: Message;
+}
+
+export function MessageBubble({ message }: Props) {
+  if (message.role === 'tool') {
+    return <ToolResultBubble message={message} />;
+  }
+
+  const isUser = message.role === 'user';
+  const isStreaming = message.status === 'streaming';
+  const isFailed = message.status === 'failed';
+
+  return (
+    <div className={`flex ${isUser ? 'justify-end' : 'justify-start'} mb-3`}>
+      <div
+        className={`max-w-[85%] rounded-lg px-4 py-2.5 ${
+          isUser
+            ? 'bg-blue-600 text-white'
+            : 'bg-zinc-800 text-zinc-100 border border-zinc-700'
+        }`}
+      >
+        {isFailed && (
+          <div className="flex items-center gap-1.5 text-red-400 text-xs mb-1">
+            <AlertCircle size={12} />
+            <span>Failed</span>
+          </div>
+        )}
+
+        {message.tool_calls && message.tool_calls.length > 0 && (
+          <div className="mb-2 space-y-1">
+            {message.tool_calls.map((tc) => (
+              <div
+                key={tc.id}
+                className="flex items-center gap-1.5 text-xs text-zinc-400 bg-zinc-900/50 rounded px-2 py-1"
+              >
+                <Wrench size={11} />
+                <span className="font-mono">{tc.name}</span>
+                <span className="text-zinc-500 truncate max-w-[200px]">
+                  {truncateArgs(tc.arguments)}
+                </span>
+              </div>
+            ))}
+          </div>
+        )}
+
+        {message.content.trim() && (
+          <div className="prose prose-invert prose-sm max-w-none [&_pre]:bg-zinc-900 [&_pre]:p-3 [&_pre]:rounded [&_pre]:overflow-x-auto [&_code]:text-zinc-300 [&_p]:my-1.5">
+            <Markdown remarkPlugins={[remarkGfm]}>{message.content}</Markdown>
+          </div>
+        )}
+
+        {isStreaming && !message.content.trim() && (
+          <div className="flex items-center gap-1.5 text-zinc-400">
+            <Loader2 size={14} className="animate-spin" />
+            <span className="text-xs">Thinking...</span>
+          </div>
+        )}
+
+        {isStreaming && message.content.trim() && (
+          <span className="inline-block w-1.5 h-4 bg-zinc-400 animate-pulse ml-0.5 align-middle" />
+        )}
+      </div>
+    </div>
+  );
+}
+
+function ToolResultBubble({ message }: Props) {
+  const result = message.tool_results;
+  if (!result) return null;
+
+  const isError = result.error;
+  const output = result.output || '';
+  const displayOutput =
+    output.length > 300 ? output.slice(0, 300) + '...' : output;
+
+  return (
+    <div className="flex justify-start mb-2 ml-6">
+      <div
+        className={`max-w-[80%] rounded px-3 py-2 text-xs font-mono border ${
+          isError
+            ? 'bg-red-950/30 border-red-800/50 text-red-300'
+            : 'bg-zinc-800/50 border-zinc-700/50 text-zinc-400'
+        }`}
+      >
+        {result.truncated && (
+          <span className="text-yellow-500 text-[10px] block mb-1">
+            [truncated]
+          </span>
+        )}
+        <pre className="whitespace-pre-wrap break-all">{displayOutput}</pre>
+      </div>
+    </div>
+  );
+}
+
+function truncateArgs(args: string): string {
+  if (!args) return '';
+  try {
+    const parsed = JSON.parse(args);
+    const keys = Object.keys(parsed);
+    if (keys.length === 0) return '';
+    const first = keys[0]!;
+    const val = String(parsed[first]);
+    const display = val.length > 40 ? val.slice(0, 40) + '...' : val;
+    return `${first}: ${display}`;
+  } catch {
+    return args.length > 50 ? args.slice(0, 50) + '...' : args;
+  }
+}
--- a/apps/coder/web/src/globals.css
+++ b/apps/coder/web/src/globals.css
@@ -0,0 +1,22 @@
+@import "tailwindcss";
+
+body {
+  margin: 0;
+  min-height: 100vh;
+}
+
+/* Scrollbar styling for dark theme */
+::-webkit-scrollbar {
+  width: 8px;
+  height: 8px;
+}
+::-webkit-scrollbar-track {
+  background: transparent;
+}
+::-webkit-scrollbar-thumb {
+  background: #3f3f46;
+  border-radius: 4px;
+}
+::-webkit-scrollbar-thumb:hover {
+  background: #52525b;
+}
--- a/apps/coder/web/src/hooks/useSessionStream.ts
+++ b/apps/coder/web/src/hooks/useSessionStream.ts
@@ -0,0 +1,230 @@
+import { useEffect, useRef, useState, useCallback } from 'react';
+import type { Message, WsFrame, PendingChange } from '@/api/types';
+
+interface State {
+  messages: Message[];
+  connected: boolean;
+  error: string | null;
+}
+
+function applyFrame(state: State, frame: WsFrame): State {
+  switch (frame.type) {
+    case 'snapshot': {
+      return { ...state, messages: frame.messages };
+    }
+    case 'message_started': {
+      const exists = state.messages.some((m) => m.id === frame.message_id);
+      if (exists) return state;
+      const newMsg: Message = {
+        id: frame.message_id,
+        session_id: '',
+        chat_id: frame.chat_id,
+        role: frame.role,
+        content: '',
+        kind: 'message',
+        tool_calls: null,
+        tool_results: null,
+        status: frame.role === 'system' ? 'complete' : 'streaming',
+        tokens_used: null,
+        ctx_used: null,
+        ctx_max: null,
+        started_at: null,
+        finished_at: null,
+        created_at: new Date().toISOString(),
+        metadata: null,
+      };
+      return { ...state, messages: [...state.messages, newMsg] };
+    }
+    case 'delta': {
+      const next = state.messages.map((m) =>
+        m.id === frame.message_id ? { ...m, content: m.content + frame.content } : m,
+      );
+      return { ...state, messages: next };
+    }
+    case 'tool_call': {
+      const next = state.messages.map((m) =>
+        m.id === frame.message_id
+          ? { ...m, tool_calls: [...(m.tool_calls ?? []), frame.tool_call] }
+          : m,
+      );
+      return { ...state, messages: next };
+    }
+    case 'tool_result': {
+      const exists = state.messages.some((m) => m.id === frame.tool_message_id);
+      if (exists) {
+        const next = state.messages.map((m) =>
+          m.id === frame.tool_message_id
+            ? {
+                ...m,
+                role: 'tool' as const,
+                tool_results: {
+                  tool_call_id: frame.tool_call_id,
+                  output: frame.output,
+                  truncated: frame.truncated,
+                  ...(frame.error ? { error: frame.error } : {}),
+                },
+                status: 'complete' as const,
+              }
+            : m,
+        );
+        return { ...state, messages: next };
+      }
+      const newMsg: Message = {
+        id: frame.tool_message_id,
+        session_id: '',
+        chat_id: frame.chat_id,
+        role: 'tool',
+        content: '',
+        kind: 'message',
+        tool_calls: null,
+        tool_results: {
+          tool_call_id: frame.tool_call_id,
+          output: frame.output,
+          truncated: frame.truncated,
+          ...(frame.error ? { error: frame.error } : {}),
+        },
+        status: 'complete',
+        tokens_used: null,
+        ctx_used: null,
+        ctx_max: null,
+        started_at: null,
+        finished_at: null,
+        created_at: new Date().toISOString(),
+        metadata: null,
+      };
+      return { ...state, messages: [...state.messages, newMsg] };
+    }
+    case 'message_complete': {
+      const next = state.messages.map((m) =>
+        m.id === frame.message_id
+          ? {
+              ...m,
+              status: 'complete' as const,
+              ...(frame.tokens_used !== undefined ? { tokens_used: frame.tokens_used } : {}),
+              ...(frame.ctx_used !== undefined ? { ctx_used: frame.ctx_used } : {}),
+              ...(frame.ctx_max !== undefined ? { ctx_max: frame.ctx_max } : {}),
+              ...(frame.started_at !== undefined ? { started_at: frame.started_at } : {}),
+              ...(frame.finished_at !== undefined ? { finished_at: frame.finished_at } : {}),
+              ...(frame.metadata !== undefined ? { metadata: frame.metadata } : {}),
+            }
+          : m,
+      );
+      return { ...state, messages: next };
+    }
+    case 'error': {
+      const next = frame.message_id
+        ? state.messages.map((m) =>
+            m.id === frame.message_id ? { ...m, status: 'failed' as const } : m,
+          )
+        : state.messages;
+      return { ...state, messages: next, error: frame.error };
+    }
+    case 'pending_change_added':
+    case 'pending_change_updated':
+      // These are handled by the pending changes listener, not the message state
+      return state;
+  }
+}
+
+const RECONNECT_INITIAL_MS = 1000;
+const RECONNECT_MAX_MS = 30_000;
+
+interface SessionStreamResult {
+  messages: Message[];
+  connected: boolean;
+  error: string | null;
+  isStreaming: boolean;
+  /** Listeners for pending change frames */
+  onPendingChange: (cb: (change: PendingChange) => void) => () => void;
+}
+
+export function useSessionStream(sessionId: string | undefined): SessionStreamResult {
+  const [state, setState] = useState<State>({ messages: [], connected: false, error: null });
+  const wsRef = useRef<WebSocket | null>(null);
+  const pendingListenersRef = useRef<Set<(change: PendingChange) => void>>(new Set());
+
+  useEffect(() => {
+    if (!sessionId) return;
+
+    setState({ messages: [], connected: false, error: null });
+
+    let unmounted = false;
+    let reconnectTimer: ReturnType<typeof setTimeout> | null = null;
+    let reconnectDelay = RECONNECT_INITIAL_MS;
+
+    const connect = () => {
+      if (unmounted) return;
+      const proto = window.location.protocol === 'https:' ? 'wss' : 'ws';
+      const url = `${proto}://${window.location.host}/api/ws/sessions/${sessionId}`;
+      const ws = new WebSocket(url);
+      wsRef.current = ws;
+
+      ws.onopen = () => {
+        reconnectDelay = RECONNECT_INITIAL_MS;
+        setState((s) => ({ ...s, connected: true, error: null }));
+      };
+
+      ws.onmessage = (ev) => {
+        let frame: WsFrame;
+        try {
+          frame = JSON.parse(typeof ev.data === 'string' ? ev.data : '') as WsFrame;
+        } catch {
+          return;
+        }
+
+        // Notify pending change listeners
+        if (frame.type === 'pending_change_added' || frame.type === 'pending_change_updated') {
+          for (const cb of pendingListenersRef.current) {
+            cb(frame.change);
+          }
+        }
+
+        setState((s) => applyFrame(s, frame));
+      };
+
+      ws.onerror = () => {
+        try {
+          ws.close();
+        } catch {}
+      };
+
+      ws.onclose = () => {
+        if (unmounted) return;
+        setState((s) => ({ ...s, connected: false }));
+        const delay = reconnectDelay;
+        reconnectDelay = Math.min(reconnectDelay * 2, RECONNECT_MAX_MS);
+        reconnectTimer = setTimeout(connect, delay);
+      };
+    };
+
+    connect();
+
+    return () => {
+      unmounted = true;
+      if (reconnectTimer) clearTimeout(reconnectTimer);
+      const ws = wsRef.current;
+      wsRef.current = null;
+      if (ws)
+        try {
+          ws.close();
+        } catch {}
+    };
+  }, [sessionId]);
+
+  const isStreaming = state.messages.some((m) => m.status === 'streaming');
+
+  const onPendingChange = useCallback((cb: (change: PendingChange) => void) => {
+    pendingListenersRef.current.add(cb);
+    return () => {
+      pendingListenersRef.current.delete(cb);
+    };
+  }, []);
+
+  return {
+    messages: state.messages,
+    connected: state.connected,
+    error: state.error,
+    isStreaming,
+    onPendingChange,
+  };
+}
--- a/apps/coder/web/src/main.tsx
+++ b/apps/coder/web/src/main.tsx
@@ -0,0 +1,13 @@
+import { StrictMode } from 'react';
+import { createRoot } from 'react-dom/client';
+import { BrowserRouter } from 'react-router-dom';
+import { App } from './App';
+import './globals.css';
+
+createRoot(document.getElementById('root')!).render(
+  <StrictMode>
+    <BrowserRouter>
+      <App />
+    </BrowserRouter>
+  </StrictMode>,
+);
--- a/apps/coder/web/src/pages/Home.tsx
+++ b/apps/coder/web/src/pages/Home.tsx
@@ -0,0 +1,138 @@
+import { useEffect, useState } from 'react';
+import { useNavigate } from 'react-router-dom';
+import { Code2, Folder, ArrowRight } from 'lucide-react';
+import type { Project, Session } from '@/api/types';
+import { api } from '@/api/client';
+
+export function Home() {
+  const navigate = useNavigate();
+  const [projects, setProjects] = useState<Project[]>([]);
+  const [sessions, setSessions] = useState<Session[]>([]);
+  const [selectedProject, setSelectedProject] = useState<string | null>(null);
+  const [loading, setLoading] = useState(true);
+
+  // Fetch projects on mount
+  useEffect(() => {
+    api.projects
+      .list({ status: 'open' })
+      .then(setProjects)
+      .catch(console.error)
+      .finally(() => setLoading(false));
+  }, []);
+
+  // Fetch sessions when a project is selected
+  useEffect(() => {
+    if (!selectedProject) {
+      setSessions([]);
+      return;
+    }
+    api.sessions
+      .listForProject(selectedProject, 'open')
+      .then(setSessions)
+      .catch(console.error);
+  }, [selectedProject]);
+
+  const handleSessionClick = (session: Session) => {
+    navigate(`/sessions/${session.id}`);
+  };
+
+  if (loading) {
+    return (
+      <div className="min-h-screen bg-zinc-900 flex items-center justify-center">
+        <div className="text-zinc-500">Loading...</div>
+      </div>
+    );
+  }
+
+  return (
+    <div className="min-h-screen bg-zinc-900 p-6">
+      <div className="max-w-2xl mx-auto">
+        {/* Header */}
+        <div className="flex items-center gap-3 mb-8">
+          <Code2 size={28} className="text-blue-400" />
+          <h1 className="text-2xl font-bold text-zinc-100">BooCoder</h1>
+        </div>
+
+        {/* Projects list */}
+        <div className="mb-8">
+          <h2 className="text-sm font-medium text-zinc-400 uppercase tracking-wide mb-3">
+            Projects
+          </h2>
+          {projects.length === 0 ? (
+            <p className="text-zinc-500 text-sm">
+              No projects found. Create one in BooChat first.
+            </p>
+          ) : (
+            <div className="space-y-1">
+              {projects.map((project) => (
+                <button
+                  key={project.id}
+                  onClick={() => setSelectedProject(project.id)}
+                  className={`w-full flex items-center gap-3 px-4 py-3 rounded-lg text-left transition-colors ${
+                    selectedProject === project.id
+                      ? 'bg-blue-600/20 border border-blue-500/40'
+                      : 'bg-zinc-800/50 border border-zinc-800 hover:bg-zinc-800'
+                  }`}
+                >
+                  <Folder
+                    size={16}
+                    className={
+                      selectedProject === project.id
+                        ? 'text-blue-400'
+                        : 'text-zinc-500'
+                    }
+                  />
+                  <div className="flex-1 min-w-0">
+                    <div className="text-sm font-medium text-zinc-200 truncate">
+                      {project.name}
+                    </div>
+                    <div className="text-xs text-zinc-500 truncate">
+                      {project.path}
+                    </div>
+                  </div>
+                </button>
+              ))}
+            </div>
+          )}
+        </div>
+
+        {/* Sessions list */}
+        {selectedProject && (
+          <div>
+            <h2 className="text-sm font-medium text-zinc-400 uppercase tracking-wide mb-3">
+              Sessions
+            </h2>
+            {sessions.length === 0 ? (
+              <p className="text-zinc-500 text-sm">
+                No open sessions. Create one in BooChat first.
+              </p>
+            ) : (
+              <div className="space-y-1">
+                {sessions.map((session) => (
+                  <button
+                    key={session.id}
+                    onClick={() => handleSessionClick(session)}
+                    className="w-full flex items-center gap-3 px-4 py-3 rounded-lg bg-zinc-800/50 border border-zinc-800 hover:bg-zinc-800 text-left transition-colors group"
+                  >
+                    <div className="flex-1 min-w-0">
+                      <div className="text-sm font-medium text-zinc-200 truncate">
+                        {session.name || 'Untitled session'}
+                      </div>
+                      <div className="text-xs text-zinc-500">
+                        {new Date(session.updated_at).toLocaleDateString()}
+                      </div>
+                    </div>
+                    <ArrowRight
+                      size={16}
+                      className="text-zinc-600 group-hover:text-zinc-400 transition-colors"
+                    />
+                  </button>
+                ))}
+              </div>
+            )}
+          </div>
+        )}
+      </div>
+    </div>
+  );
+}
--- a/apps/coder/web/src/pages/Session.tsx
+++ b/apps/coder/web/src/pages/Session.tsx
@@ -0,0 +1,86 @@
+import { useEffect, useState } from 'react';
+import { useParams, useNavigate } from 'react-router-dom';
+import { ArrowLeft } from 'lucide-react';
+import type { Chat } from '@/api/types';
+import { api } from '@/api/client';
+import { useSessionStream } from '@/hooks/useSessionStream';
+import { ChatPane } from '@/components/ChatPane';
+import { DiffPane } from '@/components/DiffPane';
+import { Layout } from '@/components/Layout';
+
+export function Session() {
+  const { sessionId } = useParams<{ sessionId: string }>();
+  const navigate = useNavigate();
+  const [chat, setChat] = useState<Chat | null>(null);
+  const [loading, setLoading] = useState(true);
+
+  const { messages, connected, isStreaming, onPendingChange } =
+    useSessionStream(sessionId);
+
+  // Get or create a chat for this session
+  useEffect(() => {
+    if (!sessionId) return;
+
+    api.chats
+      .listForSession(sessionId)
+      .then((chats) => {
+        // Use the first open chat, or create one
+        const openChat = chats.find((c) => c.status === 'open');
+        if (openChat) {
+          setChat(openChat);
+        } else {
+          // Create a new chat
+          return api.chats.create(sessionId).then((newChat) => {
+            setChat(newChat);
+          });
+        }
+      })
+      .catch(console.error)
+      .finally(() => setLoading(false));
+  }, [sessionId]);
+
+  if (!sessionId) {
+    navigate('/');
+    return null;
+  }
+
+  if (loading) {
+    return (
+      <div className="min-h-screen bg-zinc-900 flex items-center justify-center">
+        <div className="text-zinc-500">Loading session...</div>
+      </div>
+    );
+  }
+
+  if (!chat) {
+    return (
+      <div className="min-h-screen bg-zinc-900 flex flex-col items-center justify-center gap-4">
+        <div className="text-zinc-500">Could not load chat for this session.</div>
+        <button
+          onClick={() => navigate('/')}
+          className="flex items-center gap-2 text-sm text-blue-400 hover:text-blue-300"
+        >
+          <ArrowLeft size={14} />
+          Back to projects
+        </button>
+      </div>
+    );
+  }
+
+  return (
+    <Layout
+      chatPane={
+        <ChatPane
+          sessionId={sessionId}
+          chatId={chat.id}
+          messages={messages}
+          isStreaming={isStreaming}
+          connected={connected}
+        />
+      }
+      diffPane={
+        <DiffPane sessionId={sessionId} onPendingChange={onPendingChange} />
+      }
+    />
+  );
+}
--- a/apps/coder/web/src/vite-env.d.ts
+++ b/apps/coder/web/src/vite-env.d.ts
@@ -0,0 +1 @@
+/// <reference types="vite/client" />
--- a/apps/coder/web/tsconfig.app.json
+++ b/apps/coder/web/tsconfig.app.json
@@ -0,0 +1,27 @@
+{
+  "compilerOptions": {
+    "target": "ES2022",
+    "module": "ESNext",
+    "moduleResolution": "Bundler",
+    "strict": true,
+    "esModuleInterop": true,
+    "skipLibCheck": true,
+    "forceConsistentCasingInFileNames": true,
+    "resolveJsonModule": true,
+    "allowSyntheticDefaultImports": true,
+    "isolatedModules": true,
+    "noUncheckedIndexedAccess": true,
+    "composite": true,
+    "tsBuildInfoFile": "./node_modules/.tmp/tsconfig.app.tsbuildinfo",
+    "lib": ["ES2022", "DOM", "DOM.Iterable"],
+    "jsx": "react-jsx",
+    "noEmit": true,
+    "useDefineForClassFields": true,
+    "allowImportingTsExtensions": true,
+    "baseUrl": ".",
+    "paths": {
+      "@/*": ["./src/*"]
+    }
+  },
+  "include": ["src"]
+}
--- a/apps/coder/web/tsconfig.json
+++ b/apps/coder/web/tsconfig.json
@@ -0,0 +1,13 @@
+{
+  "files": [],
+  "references": [
+    { "path": "./tsconfig.app.json" },
+    { "path": "./tsconfig.node.json" }
+  ],
+  "compilerOptions": {
+    "baseUrl": ".",
+    "paths": {
+      "@/*": ["./src/*"]
+    }
+  }
+}
--- a/apps/coder/web/tsconfig.node.json
+++ b/apps/coder/web/tsconfig.node.json
@@ -0,0 +1,14 @@
+{
+  "compilerOptions": {
+    "target": "ES2022",
+    "module": "ESNext",
+    "moduleResolution": "Bundler",
+    "strict": true,
+    "esModuleInterop": true,
+    "skipLibCheck": true,
+    "composite": true,
+    "tsBuildInfoFile": "./node_modules/.tmp/tsconfig.node.tsbuildinfo",
+    "allowSyntheticDefaultImports": true
+  },
+  "include": ["vite.config.ts"]
+}
--- a/apps/coder/web/vite.config.d.ts
+++ b/apps/coder/web/vite.config.d.ts
@@ -0,0 +1,2 @@
+declare const _default: import("vite").UserConfig;
+export default _default;
--- a/apps/coder/web/vite.config.js
+++ b/apps/coder/web/vite.config.js
@@ -0,0 +1,25 @@
+import { defineConfig } from 'vite';
+import react from '@vitejs/plugin-react';
+import path from 'node:path';
+export default defineConfig({
+    plugins: [react()],
+    resolve: {
+        alias: {
+            '@': path.resolve(__dirname, './src'),
+        },
+    },
+    server: {
+        port: 5174,
+        proxy: {
+            '/api': {
+                target: 'http://127.0.0.1:3000',
+                changeOrigin: true,
+                ws: true,
+            },
+        },
+    },
+    build: {
+        outDir: 'dist',
+        emptyOutDir: true,
+    },
+});
--- a/apps/coder/web/vite.config.ts
+++ b/apps/coder/web/vite.config.ts
@@ -0,0 +1,26 @@
+import { defineConfig } from 'vite';
+import react from '@vitejs/plugin-react';
+import path from 'node:path';
+
+export default defineConfig({
+  plugins: [react()],
+  resolve: {
+    alias: {
+      '@': path.resolve(__dirname, './src'),
+    },
+  },
+  server: {
+    port: 5174,
+    proxy: {
+      '/api': {
+        target: 'http://127.0.0.1:3000',
+        changeOrigin: true,
+        ws: true,
+      },
+    },
+  },
+  build: {
+    outDir: 'dist',
+    emptyOutDir: true,
+  },
+});
--- a/apps/server/package.json
+++ b/apps/server/package.json
@@ -4,6 +4,23 @@
  "private": true,
  "type": "module",
  "main": "dist/index.js",
+  "exports": {
+    ".": { "types": "./dist/index.d.ts", "default": "./dist/index.js" },
+    "./inference": { "types": "./dist/services/inference/index.d.ts", "default": "./dist/services/inference/index.js" },
+    "./tools": { "types": "./dist/services/tools.d.ts", "default": "./dist/services/tools.js" },
+    "./broker": { "types": "./dist/services/broker.d.ts", "default": "./dist/services/broker.js" },
+    "./compaction": { "types": "./dist/services/compaction.d.ts", "default": "./dist/services/compaction.js" },
+    "./model-context": { "types": "./dist/services/model-context.d.ts", "default": "./dist/services/model-context.js" },
+    "./system-prompt": { "types": "./dist/services/system-prompt.d.ts", "default": "./dist/services/system-prompt.js" },
+    "./agents": { "types": "./dist/services/agents.d.ts", "default": "./dist/services/agents.js" },
+    "./truncate": { "types": "./dist/services/truncate.d.ts", "default": "./dist/services/truncate.js" },
+    "./path-guard": { "types": "./dist/services/path_guard.d.ts", "default": "./dist/services/path_guard.js" },
+    "./file-ops": { "types": "./dist/services/file_ops.d.ts", "default": "./dist/services/file_ops.js" },
+    "./types": { "types": "./dist/types/api.d.ts", "default": "./dist/types/api.js" },
+    "./ws-frames": { "types": "./dist/types/ws-frames.d.ts", "default": "./dist/types/ws-frames.js" },
+    "./db": { "types": "./dist/db.d.ts", "default": "./dist/db.js" },
+    "./config": { "types": "./dist/config.d.ts", "default": "./dist/config.js" }
+  },
  "scripts": {
    "dev": "tsx watch src/index.ts",
    "build": "tsc && node -e \"import('node:fs').then(fs=>fs.copyFileSync('src/schema.sql','dist/schema.sql'))\"",
@@ -14,6 +31,7 @@
    "@ai-sdk/openai-compatible": "^2.0.47",
    "@fastify/static": "^7.0.4",
    "@fastify/websocket": "^10.0.1",
+    "@modelcontextprotocol/sdk": "^1.29.0",
    "ai": "^6.0.190",
    "fastify": "^4.28.1",
    "postgres": "^3.4.4",
--- a/apps/server/src/config.ts
+++ b/apps/server/src/config.ts
@@ -19,6 +19,12 @@ const ConfigSchema = z.object({
  GITEA_USER: z.string().default('indifferentketchup'),
  GITEA_TOKEN: z.string().optional(),
  GITEA_SSH_HOST: z.string().default('100.114.205.53:2222'),
+  // v1.15.0-mcp-multi: path to the MCP config JSON file. Default /data/mcp.json
+  // (bind-mounted alongside AGENTS.md). File missing = no MCP (opt-in).
+  MCP_CONFIG_PATH: z.string().optional(),
+  // v2.0.5: cheaper model for titles, summaries, labeling. Falls back to
+  // session model (auto_name) or DEFAULT_MODEL when unset.
+  FAST_MODEL: z.string().optional(),
 });

 export type Config = z.infer<typeof ConfigSchema>;
--- a/apps/server/src/index.ts
+++ b/apps/server/src/index.ts
@@ -10,6 +10,7 @@ import { registerProjectRoutes } from './routes/projects.js';
 import { registerSessionRoutes } from './routes/sessions.js';
 import { registerSettingsRoutes } from './routes/settings.js';
 import { registerMessageRoutes } from './routes/messages.js';
+import { registerArtifactRoutes } from './routes/artifacts.js';
 import { registerChatRoutes } from './routes/chats.js';
 import { registerSidebarRoutes } from './routes/sidebar.js';
 import { registerWebSocket } from './routes/ws.js';
@@ -23,6 +24,10 @@ import { listSkills } from './services/skills.js';
 import * as compaction from './services/compaction.js';
 import { configureModelContext } from './services/model-context.js';
 import { cleanupTruncations } from './services/truncate.js';
+import { loadMcpConfig } from './services/mcp-config.js';
+import { initialize as initMcp, getTools as getMcpTools, shutdown as shutdownMcp } from './services/mcp-client.js';
+import { appendMcpTools } from './services/tools.js';
+import { refreshToolNames } from './services/agents.js';

 async function main() {
  const config = loadConfig();
@@ -68,6 +73,23 @@ async function main() {
  // default_generation_settings.n_ctx — the value persisted as messages.ctx_max.
  configureModelContext({ llamaSwapUrl: config.LLAMA_SWAP_URL });

+  // v1.15.0-mcp-multi: read MCP config file and connect to all enabled servers.
+  // Runs before route registration so the tool list is complete when the first
+  // inference request arrives. Per-server graceful degradation: one failing
+  // server doesn't block others.
+  const mcpConfigPath = config.MCP_CONFIG_PATH ?? '/data/mcp.json';
+  const mcpServers = loadMcpConfig(mcpConfigPath, app.log);
+  if (mcpServers.length > 0) {
+    await initMcp(mcpServers, app.log);
+    const mcpTools = getMcpTools();
+    if (mcpTools.length > 0) {
+      appendMcpTools(mcpTools);
+      refreshToolNames();
+      app.log.info({ servers: mcpServers.length, tools: mcpTools.length }, 'mcp: registered');
+    }
+  }
+  app.addHook('onClose', async () => { await shutdownMcp(); });
+
  await app.register(fastifyWebsocket);

  app.get('/api/health', async () => {
@@ -115,7 +137,7 @@ async function main() {
      broker.publishUserFrame(user, frame as unknown as import('./types/ws-frames.js').WsFrame);
    }
  );
-  registerMessageRoutes(app, sql, {
+  registerMessageRoutes(app, sql, config, broker, {
    enqueueInference: (sessionId, chatId, assistantId, user) => {
      inference.enqueue(sessionId, chatId, assistantId, user);
    },
@@ -160,6 +182,7 @@ async function main() {
      broker.publishFrame(sessionId, frame as import('./types/ws-frames.js').WsFrame);
    },
  });
+  registerArtifactRoutes(app, sql);
  registerSkillsRoutes(app, sql, {
    enqueueInference: (sessionId, chatId, assistantId, user) => {
      inference.enqueue(sessionId, chatId, assistantId, user);
@@ -189,6 +212,37 @@ async function main() {
  });
  registerWebSocket(app, sql, broker);

+  // v2.0.0: reverse proxy /api/coder/* to the boocoder container. Keeps the
+  // SPA's HTTP requests going through a single origin (avoids CORS). WS for
+  // the coder pane connects directly to boocoder:9502 from the browser (same
+  // Tailscale network — no CORS issue for WebSocket upgrade requests).
+  const BOOCODER_ORIGIN = process.env.BOOCODER_URL ?? 'http://boocoder:3000';
+  app.all('/api/coder/*', async (req, reply) => {
+    const targetPath = req.url.replace('/api/coder', '/api');
+    const targetUrl = `${BOOCODER_ORIGIN}${targetPath}`;
+    const headers: Record<string, string> = {};
+    if (req.headers['content-type']) headers['content-type'] = req.headers['content-type'] as string;
+    if (req.headers['authorization']) headers['authorization'] = req.headers['authorization'] as string;
+
+    try {
+      const res = await fetch(targetUrl, {
+        method: req.method as string,
+        headers,
+        body: req.method !== 'GET' && req.method !== 'HEAD' ? JSON.stringify(req.body) : undefined,
+      });
+      reply.code(res.status);
+      for (const [key, value] of res.headers) {
+        if (key === 'transfer-encoding') continue;
+        reply.header(key, value);
+      }
+      const body = await res.text();
+      return reply.send(body);
+    } catch (err) {
+      app.log.error({ err, targetUrl }, 'coder proxy error');
+      reply.code(502).send({ error: 'boocoder backend unavailable' });
+    }
+  });
+
  const webDist = process.env.WEB_DIST_PATH ?? resolve(process.cwd(), '../web/dist');
  if (existsSync(webDist)) {
    await app.register(fastifyStatic, {
--- a/apps/server/src/routes/tests/sessions.test.ts
+++ b/apps/server/src/routes/tests/sessions.test.ts
@@ -0,0 +1,70 @@
+// v1.13.17-cross-repo-reads: PATCH /api/sessions/:id allowed_read_paths
+// subset enforcement. Sam flagged in the compliance review that without a
+// runtime subset check, a malicious client could POST
+//   {"allowed_read_paths":["/etc"]}
+// and bypass the user-consent grant flow entirely. The findUnauthorizedAdditions
+// helper is the guard; tests pin its behavior so a regression in the helper
+// or its callsite (PATCH handler in sessions.ts) trips CI before prod.
+
+import { describe, it, expect } from 'vitest';
+import { findUnauthorizedAdditions } from '../sessions.js';
+
+describe('findUnauthorizedAdditions — PATCH allowed_read_paths subset guard', () => {
+  it('returns no extras when requested is empty (full revoke)', () => {
+    expect(findUnauthorizedAdditions(['/opt/forks/foo'], [])).toEqual([]);
+  });
+
+  it('returns no extras when requested is a strict subset (single revoke)', () => {
+    expect(
+      findUnauthorizedAdditions(['/opt/forks/foo', '/opt/forks/bar'], ['/opt/forks/foo']),
+    ).toEqual([]);
+  });
+
+  it('returns no extras when requested equals prior (no-op PATCH)', () => {
+    expect(
+      findUnauthorizedAdditions(['/opt/forks/foo', '/opt/forks/bar'], [
+        '/opt/forks/foo',
+        '/opt/forks/bar',
+      ]),
+    ).toEqual([]);
+  });
+
+  it('flags an unauthorized addition when prior is empty', () => {
+    // The /etc bypass attempt — Sam's specific concern from the compliance
+    // review. Without this guard, the PATCH would have written /etc directly.
+    expect(findUnauthorizedAdditions([], ['/etc'])).toEqual(['/etc']);
+  });
+
+  it('flags a single unauthorized addition mixed in with valid revokes', () => {
+    // The attacker still tries to be sneaky: keep one legit entry, drop
+    // another, slip in a new one. The guard catches the addition regardless
+    // of how the rest of the array shrinks.
+    expect(
+      findUnauthorizedAdditions(['/opt/forks/foo', '/opt/forks/bar'], [
+        '/opt/forks/foo',
+        '/var/secrets',
+      ]),
+    ).toEqual(['/var/secrets']);
+  });
+
+  it('flags every unauthorized addition when there are multiple', () => {
+    expect(
+      findUnauthorizedAdditions(['/opt/forks/foo'], ['/opt/forks/foo', '/etc', '/root']),
+    ).toEqual(['/etc', '/root']);
+  });
+
+  it('treats requested duplicates correctly (each occurrence checked)', () => {
+    // If the requested array has duplicates of an unauthorized entry, the
+    // guard surfaces each one. (A frontend would never send duplicates, but
+    // the guard's contract shouldn't assume that.)
+    expect(findUnauthorizedAdditions([], ['/etc', '/etc'])).toEqual(['/etc', '/etc']);
+  });
+
+  it('does not flag entries present in prior even if requested has duplicates', () => {
+    // Duplicate of an authorized entry passes — the membership check is by
+    // value, not by index. Settled by Set.has semantics.
+    expect(
+      findUnauthorizedAdditions(['/opt/forks/foo'], ['/opt/forks/foo', '/opt/forks/foo']),
+    ).toEqual([]);
+  });
+});
--- a/apps/server/src/routes/artifacts.ts
+++ b/apps/server/src/routes/artifacts.ts
@@ -0,0 +1,231 @@
+// v1.14.x-html-artifact-panes: artifact download routes.
+//
+// Two endpoints:
+//   POST /api/chats/:id/messages/:msg_id/artifacts/download?fmt=md|html
+//     Materialises a file under <projectRoot>/.boocode/artifacts/ and
+//     returns {path, url}. fmt=html requires an existing html_artifact part
+//     on the message (404 otherwise). fmt=md works on any assistant
+//     message with non-empty content.
+//
+//   GET /api/projects/:project_id/artifacts/:filename
+//     Streams a previously-written artifact back with
+//     Content-Disposition: attachment. Path-guarded to the project's
+//     artifacts dir; rejects traversal attempts.
+
+import { createReadStream } from 'node:fs';
+import { realpath, stat } from 'node:fs/promises';
+import { resolve, sep, basename } from 'node:path';
+import type { FastifyInstance } from 'fastify';
+import { z } from 'zod';
+import type { Sql } from '../db.js';
+import {
+  writeHtmlArtifact,
+  writeMarkdownArtifact,
+  type HtmlArtifactPayload,
+} from '../services/artifacts.js';
+
+const DownloadQuery = z.object({
+  fmt: z.enum(['md', 'html']),
+});
+
+// Filename safety: alnum, dash, dot, underscore only. Blocks `..`, slashes,
+// nul bytes, etc. before we even touch the filesystem.
+const FilenameRe = /^[A-Za-z0-9._-]+$/;
+
+interface ChatRow {
+  id: string;
+  session_id: string;
+  project_id: string;
+  project_path: string;
+}
+
+interface MessageRow {
+  id: string;
+  chat_id: string;
+  role: string;
+  content: string;
+}
+
+export function registerArtifactRoutes(app: FastifyInstance, sql: Sql): void {
+  app.post<{
+    Params: { id: string; msg_id: string };
+    Querystring: { fmt?: string };
+  }>(
+    '/api/chats/:id/messages/:msg_id/artifacts/download',
+    async (req, reply) => {
+      const parsed = DownloadQuery.safeParse(req.query);
+      if (!parsed.success) {
+        reply.code(400);
+        return { error: 'invalid query', details: parsed.error.flatten() };
+      }
+      const { fmt } = parsed.data;
+      const { id: chatId, msg_id: messageId } = req.params;
+
+      const chatRows = await sql<ChatRow[]>`
+        SELECT c.id, c.session_id, s.project_id, p.path AS project_path
+        FROM chats c
+        JOIN sessions s ON s.id = c.session_id
+        JOIN projects p ON p.id = s.project_id
+        WHERE c.id = ${chatId}
+      `;
+      if (chatRows.length === 0) {
+        reply.code(404);
+        return { error: 'chat not found' };
+      }
+      const chat = chatRows[0]!;
+
+      const msgRows = await sql<MessageRow[]>`
+        SELECT id, chat_id, role, content
+        FROM messages
+        WHERE id = ${messageId} AND chat_id = ${chatId}
+      `;
+      if (msgRows.length === 0) {
+        reply.code(404);
+        return { error: 'message not found' };
+      }
+      const msg = msgRows[0]!;
+      if (msg.role !== 'assistant') {
+        reply.code(400);
+        return { error: 'only assistant messages produce artifacts' };
+      }
+
+      const ctx = { projectId: chat.project_id, projectRoot: chat.project_path };
+
+      try {
+        if (fmt === 'md') {
+          if (!msg.content || msg.content.trim().length === 0) {
+            reply.code(400);
+            return { error: 'message has no content to export' };
+          }
+          const result = await writeMarkdownArtifact(
+            { content: msg.content },
+            ctx,
+          );
+          return result;
+        }
+        // fmt === 'html': require an html_artifact part on the message.
+        const partRows = await sql<{ payload: HtmlArtifactPayload }[]>`
+          SELECT payload
+          FROM message_parts
+          WHERE message_id = ${messageId} AND kind = 'html_artifact'
+          ORDER BY sequence ASC
+          LIMIT 1
+        `;
+        if (partRows.length === 0) {
+          reply.code(404);
+          return { error: 'no html_artifact part on this message' };
+        }
+        const result = await writeHtmlArtifact(partRows[0]!.payload, ctx);
+        return result;
+      } catch (err) {
+        req.log.error({ err, messageId, fmt }, 'artifact write failed');
+        reply.code(500);
+        return {
+          error: err instanceof Error ? err.message : 'artifact write failed',
+        };
+      }
+    },
+  );
+
+  // v1.14.x-html-artifact-panes: HtmlArtifactPane needs the payload on click
+  // to render its iframe. Returns 404 when the message has no html_artifact
+  // sibling part — frontend uses that signal to open the markdown_artifact
+  // pane variant instead. Payload shape matches HtmlArtifactPayload in
+  // services/artifacts.ts.
+  app.get<{ Params: { id: string; msg_id: string } }>(
+    '/api/chats/:id/messages/:msg_id/html_artifact',
+    async (req, reply) => {
+      const { id: chatId, msg_id: messageId } = req.params;
+      const partRows = await sql<{ payload: HtmlArtifactPayload }[]>`
+        SELECT payload
+        FROM message_parts mp
+        JOIN messages m ON m.id = mp.message_id
+        WHERE mp.message_id = ${messageId}
+          AND m.chat_id = ${chatId}
+          AND mp.kind = 'html_artifact'
+        ORDER BY mp.sequence ASC
+        LIMIT 1
+      `;
+      if (partRows.length === 0) {
+        reply.code(404);
+        return { error: 'no html_artifact part on this message' };
+      }
+      return partRows[0]!.payload;
+    },
+  );
+
+  app.get<{ Params: { project_id: string; filename: string } }>(
+    '/api/projects/:project_id/artifacts/:filename',
+    async (req, reply) => {
+      const { project_id: projectId, filename } = req.params;
+      // Strip directory components defensively; only the basename is allowed.
+      const base = basename(filename);
+      if (base !== filename || !FilenameRe.test(base)) {
+        reply.code(400);
+        return { error: 'invalid filename' };
+      }
+
+      const projectRows = await sql<{ id: string; path: string }[]>`
+        SELECT id, path FROM projects WHERE id = ${projectId}
+      `;
+      if (projectRows.length === 0) {
+        reply.code(404);
+        return { error: 'project not found' };
+      }
+      const project = projectRows[0]!;
+
+      let resolvedRoot: string;
+      try {
+        resolvedRoot = await realpath(project.path);
+      } catch {
+        reply.code(404);
+        return { error: 'project path missing' };
+      }
+      const artifactsDir = resolve(resolvedRoot, '.boocode/artifacts');
+      const absPath = resolve(artifactsDir, base);
+      if (!absPath.startsWith(artifactsDir + sep)) {
+        reply.code(400);
+        return { error: 'path traversal rejected' };
+      }
+      // Close the symlink-escape gap: if `.boocode/artifacts` (or an
+      // ancestor) is a symlink pointing outside resolvedRoot, the lexical
+      // prefix check above passes but the actual read lands outside the
+      // sandbox. Realpath the artifacts dir and re-verify.
+      try {
+        const realArtifactsDir = await realpath(artifactsDir);
+        if (
+          realArtifactsDir !== resolvedRoot &&
+          !realArtifactsDir.startsWith(resolvedRoot + sep)
+        ) {
+          reply.code(400);
+          return { error: 'path traversal rejected' };
+        }
+      } catch {
+        reply.code(404);
+        return { error: 'artifact not found' };
+      }
+      try {
+        await stat(absPath);
+      } catch {
+        reply.code(404);
+        return { error: 'artifact not found' };
+      }
+      const ext = base.toLowerCase().endsWith('.html')
+        ? 'text/html; charset=utf-8'
+        : base.toLowerCase().endsWith('.md')
+          ? 'text/markdown; charset=utf-8'
+          : 'application/octet-stream';
+      reply.header('Content-Type', ext);
+      // Defense-in-depth on LLM-generated HTML served through this route.
+      // Authelia gates the proxy; these headers limit blast radius if a
+      // payload tries to escape that boundary in-browser.
+      reply.header('X-Content-Type-Options', 'nosniff');
+      reply.header('Content-Security-Policy', 'sandbox');
+      reply.header(
+        'Content-Disposition',
+        `attachment; filename="${base.replace(/"/g, '')}"`,
+      );
+      return reply.send(createReadStream(absPath));
+    },
+  );
+}
--- a/apps/server/src/routes/chats.ts
+++ b/apps/server/src/routes/chats.ts
@@ -296,13 +296,13 @@ export function registerChatRoutes(
        `;
        await tx`
          INSERT INTO messages (
-            session_id, chat_id, role, content, kind, tool_calls, tool_results,
+            session_id, chat_id, role, content, kind,
            status, tokens_used, ctx_used, ctx_max, started_at, finished_at,
            created_at, metadata
          )
          SELECT
            ${source.session_id}, ${chat!.id}, role, content, kind,
-            tool_calls, tool_results, status,
+            status,
            tokens_used, ctx_used, ctx_max, started_at, finished_at,
            clock_timestamp() + (
              ROW_NUMBER() OVER (ORDER BY created_at ASC, id ASC) * INTERVAL '1 microsecond'
@@ -385,21 +385,25 @@ export function registerChatRoutes(
        reply.code(409);
        return { error: 'message is not stale yet', age_seconds: msg.age_seconds };
      }
-      const updated = await sql<Message[]>`
+      const updated = await sql<{ id: string }[]>`
        UPDATE messages
        SET status = 'failed',
            content = COALESCE(content, ''),
            finished_at = clock_timestamp()
        WHERE id = ${msg.id} AND status = 'streaming'
-        RETURNING id, session_id, chat_id, role, content, kind, tool_calls, tool_results,
-                  status, last_seq, tokens_used, ctx_used, ctx_max, started_at, finished_at,
-                  created_at, metadata, summary, tail_start_id, compacted_at
+        RETURNING id
      `;
      if (updated.length === 0) {
        // Race: the row flipped out of 'streaming' between our SELECT and UPDATE.
        reply.code(409);
        return { error: 'message status changed mid-request' };
      }
+      // v1.13.20: re-fetch via messages_with_parts so the returned shape
+      // carries parts-synthesized tool_calls / tool_results. The dropped
+      // legacy columns can no longer be selected directly.
+      const refreshed = await sql<Message[]>`
+        SELECT * FROM messages_with_parts WHERE id = ${msg.id}
+      `;
      broker.publishUserFrame('default', {
        type: 'chat_status',
        chat_id: msg.chat_id,
@@ -411,7 +415,7 @@ export function registerChatRoutes(
        message_id: msg.id,
        chat_id: msg.chat_id,
      });
-      return updated[0];
+      return refreshed[0];
    }
  );

--- a/apps/server/src/routes/messages.ts
+++ b/apps/server/src/routes/messages.ts
@@ -1,7 +1,13 @@
 import type { FastifyInstance } from 'fastify';
 import { z } from 'zod';
 import type { Sql } from '../db.js';
+import type { Config } from '../config.js';
+import type { Broker } from '../services/broker.js';
 import type { Chat, Message, Session, ToolCall } from '../types/api.js';
+// v1.13.17-cross-repo-reads: grant_read_access resolves the grant root at
+// decision time (not at request time) so concurrent project changes don't
+// stale-bind the resolution.
+import { resolveGrantRoot } from '../services/grant_resolver.js';

 const SendBody = z.object({
  content: z.string().min(1).max(64_000),
@@ -47,6 +53,21 @@ const AskUserInputArgs = z.object({
    .max(3),
 });

+// v1.13.17-cross-repo-reads: grant decision body. tool_call_id is the
+// model-emitted id (e.g. "call_abc123"), not a UUID. decision is binary.
+const GrantReadAccessBody = z.object({
+  tool_call_id: z.string().min(1),
+  decision: z.enum(['allow', 'deny']),
+});
+
+// Same shape as services/request_read_access.ts RequestReadAccessInput.
+// Re-derived to avoid the services/tools.ts import (matches the
+// AskUserInputArgs pattern above).
+const RequestReadAccessArgs = z.object({
+  path: z.string().min(1),
+  reason: z.string().min(1).max(500),
+});
+
 interface MessageHandlers {
  enqueueInference: (sessionId: string, chatId: string, assistantMessageId: string, user: string) => void;
  // v1.11: returns a promise that resolves after compaction.process finishes
@@ -76,6 +97,8 @@ interface MessageHandlers {
 export function registerMessageRoutes(
  app: FastifyInstance,
  sql: Sql,
+  config: Config,
+  broker: Broker,
  handlers: MessageHandlers
 ): void {
  app.get<{ Params: { id: string } }>(
@@ -582,15 +605,11 @@ export function registerMessageRoutes(

      const toolMessageId = toolRow.message_id;
      const result = await sql.begin(async (tx) => {
-        await tx`
-          UPDATE messages
-          SET tool_results = ${tx.json(newToolResults as never)}
-          WHERE id = ${toolMessageId}
-        `;
-        // v1.13.0: replace the pending tool_result part inserted at message
-        // creation (tool-phase.ts) with the answered one. Delete-then-insert
-        // is simpler than UPDATE because parts are append-style elsewhere;
-        // the UNIQUE (message_id, sequence) constraint blocks plain insert.
+        // v1.13.20: parts-only. Replace the pending tool_result part inserted
+        // at message creation (tool-phase.ts) with the answered one. Delete-
+        // then-insert is simpler than UPDATE because parts are append-style
+        // elsewhere; the UNIQUE (message_id, sequence) constraint blocks
+        // plain insert.
        await tx`DELETE FROM message_parts WHERE message_id = ${toolMessageId} AND kind = 'tool_result'`;
        await tx`
          INSERT INTO message_parts (message_id, sequence, kind, payload)
@@ -626,4 +645,230 @@ export function registerMessageRoutes(
      return result;
    },
  );
+
+  // v1.13.17-cross-repo-reads: resume an awaiting-grant pause. Mirror shape
+  // of /answer_user_input (validate, look up via message_parts, UPDATE,
+  // publish, enqueue). Differences vs /answer_user_input:
+  //   - On 'allow', re-resolves the grant root via grant_resolver (state
+  //     may have changed since the prompt fired — concurrent project add,
+  //     etc.). Resolution failure auto-falls to a denial with reason text
+  //     rather than 500ing.
+  //   - On 'allow' with a valid root, appends to sessions.allowed_read_paths
+  //     (deduplicated) inside the same transaction.
+  //   - On success, also publishes session_updated so an open SettingsPane
+  //     refetches the new grant list.
+  // Error codes match /answer:
+  //   400 invalid_body / mismatched_answer_shape (bad args on the tool_call)
+  //   404 chat_not_found / unknown_tool_call_id
+  //   409 tool_call_already_answered
+  app.post<{ Params: { id: string } }>(
+    '/api/chats/:id/grant_read_access',
+    async (req, reply) => {
+      const parsed = GrantReadAccessBody.safeParse(req.body);
+      if (!parsed.success) {
+        reply.code(400);
+        return { error: 'invalid_body', details: parsed.error.flatten() };
+      }
+      const { tool_call_id, decision } = parsed.data;
+
+      const chatRows = await sql<Chat[]>`
+        SELECT id, session_id FROM chats WHERE id = ${req.params.id} AND status = 'open'
+      `;
+      if (chatRows.length === 0) {
+        reply.code(404);
+        return { error: 'chat_not_found' };
+      }
+      const chat = chatRows[0]!;
+      const sessionId = chat.session_id;
+
+      // Mirror the /answer lookup: assistant tool_call by id via message_parts.
+      const callerRows = await sql<{
+        message_id: string;
+        payload: { id: string; name: string; args: Record<string, unknown> };
+      }[]>`
+        SELECT p.message_id, p.payload
+        FROM message_parts p
+        JOIN messages m ON m.id = p.message_id
+        WHERE m.chat_id = ${chat.id}
+          AND m.role = 'assistant'
+          AND p.kind = 'tool_call'
+          AND p.payload->>'id' = ${tool_call_id}
+        ORDER BY m.created_at DESC
+        LIMIT 1
+      `;
+      const callerRow = callerRows[0];
+      if (!callerRow) {
+        reply.code(404);
+        return { error: 'unknown_tool_call_id' };
+      }
+      const foundCall: ToolCall = {
+        id: callerRow.payload.id,
+        name: callerRow.payload.name,
+        args: callerRow.payload.args,
+      };
+      if (foundCall.name !== 'request_read_access') {
+        reply.code(400);
+        return { error: 'tool_call_not_request_read_access' };
+      }
+      const argsParsed = RequestReadAccessArgs.safeParse(foundCall.args);
+      if (!argsParsed.success) {
+        reply.code(400);
+        return { error: 'mismatched_answer_shape', detail: 'tool_call args invalid' };
+      }
+      const requestedPath = argsParsed.data.path;
+
+      // Find the pending tool row.
+      const toolRows = await sql<{
+        message_id: string;
+        payload: { tool_call_id: string; output: unknown };
+      }[]>`
+        SELECT p.message_id, p.payload
+        FROM message_parts p
+        JOIN messages m ON m.id = p.message_id
+        WHERE m.chat_id = ${chat.id}
+          AND m.role = 'tool'
+          AND p.kind = 'tool_result'
+          AND p.payload->>'tool_call_id' = ${tool_call_id}
+        ORDER BY m.created_at DESC
+        LIMIT 1
+      `;
+      const toolRow = toolRows[0];
+      if (!toolRow) {
+        reply.code(404);
+        return { error: 'unknown_tool_call_id', detail: 'tool message not found' };
+      }
+      if (toolRow.payload && toolRow.payload.output !== null) {
+        reply.code(409);
+        return { error: 'tool_call_already_answered' };
+      }
+
+      // Look up session + project so we can re-resolve the grant root and
+      // append to allowed_read_paths atomically. We don't need agent or
+      // history here — just the project path for the resolver.
+      const sessionRows = await sql<{
+        id: string;
+        project_id: string;
+        allowed_read_paths: string[];
+        project_path: string;
+      }[]>`
+        SELECT s.id, s.project_id, s.allowed_read_paths, p.path AS project_path
+        FROM sessions s
+        JOIN projects p ON p.id = s.project_id
+        WHERE s.id = ${sessionId}
+      `;
+      const sessionRow = sessionRows[0];
+      if (!sessionRow) {
+        reply.code(404);
+        return { error: 'session_not_found' };
+      }
+
+      // Decision branch. 'deny' is the easy path: nothing to resolve or
+      // persist. 'allow' resolves the grant root; if resolution fails (e.g.
+      // path was deleted, project removed since prompt) the tool gets a
+      // denial with the resolver's reason text instead of a 500.
+      let resultOutput: string;
+      let grantRoot: string | null = null;
+      if (decision === 'allow') {
+        const resolution = await resolveGrantRoot(
+          sql,
+          requestedPath,
+          sessionRow.project_path,
+          config.PROJECT_ROOT_WHITELIST,
+        );
+        if (!resolution.ok) {
+          resultOutput = `denied: ${resolution.reason}`;
+        } else {
+          grantRoot = resolution.root;
+          resultOutput = `granted: ${grantRoot}`;
+        }
+      } else {
+        resultOutput = 'denied';
+      }
+
+      const newToolResults = {
+        tool_call_id,
+        output: resultOutput,
+        truncated: false,
+      };
+      const toolMessageId = toolRow.message_id;
+      const dbResult = await sql.begin(async (tx) => {
+        // v1.13.20: parts-only. Same delete+insert dance as /answer —
+        // UNIQUE (message_id, sequence) blocks plain UPDATE on append-style
+        // parts.
+        await tx`DELETE FROM message_parts WHERE message_id = ${toolMessageId} AND kind = 'tool_result'`;
+        await tx`
+          INSERT INTO message_parts (message_id, sequence, kind, payload)
+          VALUES (${toolMessageId}, 0, 'tool_result', ${tx.json(newToolResults as never)})
+        `;
+        // Persist the grant if we have one. ARRAY-level dedup — append only
+        // when the root isn't already present. The session row gets
+        // touched (updated_at) so the post-update publish below has a
+        // fresh timestamp.
+        let allowedRootsAfter = sessionRow.allowed_read_paths;
+        if (grantRoot !== null) {
+          if (!sessionRow.allowed_read_paths.includes(grantRoot)) {
+            const updated = await tx<{ allowed_read_paths: string[] }[]>`
+              UPDATE sessions
+              SET allowed_read_paths = array_append(allowed_read_paths, ${grantRoot}),
+                  updated_at = clock_timestamp()
+              WHERE id = ${sessionId}
+              RETURNING allowed_read_paths
+            `;
+            allowedRootsAfter = updated[0]?.allowed_read_paths ?? sessionRow.allowed_read_paths;
+          } else {
+            // Already present — touch updated_at so any open settings
+            // panel still picks up the no-op via session_updated.
+            await tx`UPDATE sessions SET updated_at = clock_timestamp() WHERE id = ${sessionId}`;
+          }
+        }
+        const [assistantMsg] = await tx<{ id: string }[]>`
+          INSERT INTO messages (session_id, chat_id, role, content, status, created_at)
+          VALUES (${sessionId}, ${chat.id}, 'assistant', '', 'streaming', clock_timestamp())
+          RETURNING id
+        `;
+        await tx`UPDATE chats SET updated_at = clock_timestamp() WHERE id = ${chat.id}`;
+        return {
+          tool_message_id: toolMessageId,
+          assistant_message_id: assistantMsg!.id,
+          allowed_roots_after: allowedRootsAfter,
+        };
+      });
+
+      // Publish the deferred tool_result frame so the pending card flips to
+      // its answered view without a refetch.
+      handlers.publishSessionFrame(sessionId, {
+        type: 'tool_result',
+        tool_message_id: dbResult.tool_message_id,
+        tool_call_id,
+        chat_id: chat.id,
+        output: resultOutput,
+        truncated: false,
+      });
+      // session_updated nudge so any open SettingsPane refetches and sees
+      // the new allowed_read_paths. We publish on the user channel to match
+      // the existing PATCH /api/sessions/:id behavior — frontend refetches
+      // via api.sessions.get on receipt.
+      const nowIso = new Date().toISOString();
+      broker.publishUserFrame('default', {
+        type: 'session_updated',
+        session_id: sessionId,
+        project_id: sessionRow.project_id,
+        // session name doesn't change on grant; we look it up fresh to
+        // avoid carrying stale state if a rename raced us.
+        name:
+          (
+            await sql<{ name: string }[]>`SELECT name FROM sessions WHERE id = ${sessionId}`
+          )[0]?.name ?? '',
+        updated_at: nowIso,
+      });
+      handlers.enqueueInference(sessionId, chat.id, dbResult.assistant_message_id, 'default');
+
+      reply.code(202);
+      return {
+        tool_message_id: dbResult.tool_message_id,
+        assistant_message_id: dbResult.assistant_message_id,
+        allowed_read_paths: dbResult.allowed_roots_after,
+      };
+    },
+  );
 }
--- a/apps/server/src/routes/sessions.ts
+++ b/apps/server/src/routes/sessions.ts
@@ -13,12 +13,37 @@ const CreateBody = z.object({
  agent_id: z.string().min(1).max(200).nullable().optional(),
 });

+// v1.14.x-html-artifact-panes: 'markdown_artifact' + 'html_artifact' added
+// as pane kinds. Pane state is a reference only (chat_id + message_id +
+// title) — the actual artifact body is fetched from the message row or
+// message_parts.payload by the pane component on mount.
+const MarkdownArtifactStateZ = z.object({
+  chat_id: z.string().min(1).max(200),
+  message_id: z.string().min(1).max(200),
+  title: z.string().max(500),
+});
+const HtmlArtifactStateZ = z.object({
+  chat_id: z.string().min(1).max(200),
+  message_id: z.string().min(1).max(200),
+  title: z.string().max(500),
+});
+
 const WorkspacePaneZ = z.object({
  id: z.string().min(1).max(200),
-  kind: z.enum(['chat', 'terminal', 'agent', 'empty', 'settings']),
+  kind: z.enum([
+    'chat',
+    'terminal',
+    'agent',
+    'empty',
+    'settings',
+    'markdown_artifact',
+    'html_artifact',
+  ]),
  chatId: z.string().min(1).max(200).optional(),
  chatIds: z.array(z.string().min(1).max(200)).max(50),
  activeChatIdx: z.number().int(),
+  markdown_artifact_state: MarkdownArtifactStateZ.optional(),
+  html_artifact_state: HtmlArtifactStateZ.optional(),
 });

 const WorkspacePanesBody = z.object({
@@ -32,6 +57,29 @@ const PatchBody = z.object({
  agent_id: z.string().min(1).max(200).nullable().optional(),
  // v1.9: null = inherit from project default; true/false = explicit override.
  web_search_enabled: z.boolean().nullable().optional(),
+  // v1.13.17-cross-repo-reads: revocation pathway. PATCH with a shortened
+  // list deletes entries; the grant flow itself APPENDS via the separate
+  // grant_read_access endpoint, never via this PATCH. Frontend treats this
+  // as "send the new whole array". Per-entry shape validation: must be
+  // absolute, no NUL, no `/..` traversal segment. Server doesn't re-validate
+  // whitelist membership on PATCH — entries already in the array were
+  // placed there by the grant endpoint after a full whitelist+repo-shape
+  // check. THE SUBSET CHECK (every entry must already be in the current
+  // array) is enforced at runtime in the PATCH handler below, NOT in this
+  // zod refinement, because the refinement has no access to the existing
+  // session row.
+  allowed_read_paths: z
+    .array(
+      z
+        .string()
+        .min(1)
+        .max(1024)
+        .refine((p) => p.startsWith('/') && !p.includes('\0') && !p.includes('/..'), {
+          message: 'must be an absolute path without traversal markers',
+        }),
+    )
+    .max(64)
+    .optional(),
 });

 async function resolveDefaultModel(sql: Sql, config: Config): Promise<string> {
@@ -40,6 +88,19 @@ async function resolveDefaultModel(sql: Sql, config: Config): Promise<string> {
  return config.DEFAULT_MODEL;
 }

+// v1.13.17-cross-repo-reads: subset enforcement for PATCH allowed_read_paths.
+// The PATCH route can only SHRINK the array; growth happens exclusively via
+// POST /api/chats/:id/grant_read_access (which requires user consent).
+// Returns the list of disallowed-additions; an empty list means the request
+// is a valid shrink-or-no-op. Exported for the unit test.
+export function findUnauthorizedAdditions(
+  prior: readonly string[],
+  requested: readonly string[],
+): string[] {
+  const priorSet = new Set(prior);
+  return requested.filter((p) => !priorSet.has(p));
+}
+
 export function registerSessionRoutes(
  app: FastifyInstance,
  sql: Sql,
@@ -56,7 +117,7 @@ export function registerSessionRoutes(
      }
      const status = req.query.status === 'archived' ? 'archived' : 'open';
      const rows = await sql<Session[]>`
-        SELECT id, project_id, name, model, system_prompt, status, created_at, updated_at, agent_id, web_search_enabled, workspace_panes
+        SELECT id, project_id, name, model, system_prompt, status, created_at, updated_at, agent_id, web_search_enabled, workspace_panes, allowed_read_paths
        FROM sessions
        WHERE project_id = ${req.params.id} AND status = ${status}
        ORDER BY updated_at DESC
@@ -124,7 +185,7 @@ export function registerSessionRoutes(

  app.get<{ Params: { id: string } }>('/api/sessions/:id', async (req, reply) => {
    const rows = await sql<Session[]>`
-      SELECT id, project_id, name, model, system_prompt, status, created_at, updated_at, agent_id, web_search_enabled, workspace_panes
+      SELECT id, project_id, name, model, system_prompt, status, created_at, updated_at, agent_id, web_search_enabled, workspace_panes, allowed_read_paths
      FROM sessions WHERE id = ${req.params.id}
    `;
    if (rows.length === 0) {
@@ -150,15 +211,53 @@ export function registerSessionRoutes(
      const newAgentId = parsed.data.agent_id ?? null;
      const wseProvided = parsed.data.web_search_enabled !== undefined;
      const newWse = parsed.data.web_search_enabled ?? null;
-      // Read the prior name so the post-update publish can skip no-op renames
-      // (PATCH { name: "Foo" } where the session is already "Foo"). The window
-      // between SELECT and UPDATE is sub-millisecond in the same request handler;
-      // a concurrent rename in that gap would just mean one stale publish, which
-      // existing clients dedup by id.
-      const before = await sql<{ name: string }[]>`
-        SELECT name FROM sessions WHERE id = ${req.params.id}
+      // v1.13.17-cross-repo-reads: tri-state on the wire (undefined = no
+      // change, [] = clear). Frontend currently uses this PATCH only for
+      // revocation (delete a single entry from the existing array, send
+      // shortened result). Append-style grants go through the dedicated
+      // grant_read_access endpoint inside the inference loop.
+      const arpProvided = parsed.data.allowed_read_paths !== undefined;
+      const newArp = parsed.data.allowed_read_paths ?? [];
+      // Read the prior name + grants so the post-update publish can skip no-op
+      // renames (PATCH { name: "Foo" } where the session is already "Foo") AND
+      // so the subset check below has the current grant list to compare against.
+      // The window between SELECT and UPDATE is sub-millisecond in the same
+      // request handler; a concurrent rename in that gap would just mean one
+      // stale publish, which existing clients dedup by id.
+      const before = await sql<{ name: string; allowed_read_paths: string[] }[]>`
+        SELECT name, allowed_read_paths FROM sessions WHERE id = ${req.params.id}
      `;
      const priorName = before[0]?.name;
+      const priorArp = before[0]?.allowed_read_paths ?? [];
+
+      // v1.13.17-cross-repo-reads: subset enforcement. The grant flow is the
+      // ONLY path that can add entries to allowed_read_paths — PATCH can only
+      // shrink the array, never grow it. Without this guard, a malicious
+      // client could POST {"allowed_read_paths":["/etc"]} and bypass the
+      // user-consent prompt entirely. Sam flagged this in the v1.13.17
+      // compliance review (2026-05-22).
+      // Race note: a concurrent grant landing between this SELECT and the
+      // UPDATE below would briefly make a "shouldn't-have-been-valid" PATCH
+      // succeed (the newly-granted root sneaks in). Inverse race — a
+      // legitimate revoke happening alongside a concurrent grant — could
+      // briefly reject the revoke; the user retries. Both are acceptable
+      // given the single-user threat model + sub-millisecond window.
+      if (arpProvided) {
+        const extras = findUnauthorizedAdditions(priorArp, newArp);
+        if (extras.length > 0) {
+          reply.code(400);
+          return {
+            error: 'invalid body',
+            details: {
+              fieldErrors: {
+                allowed_read_paths: [
+                  `entries must already be granted; cannot add via PATCH: ${extras.join(', ')}`,
+                ],
+              },
+            },
+          };
+        }
+      }
      const rows = await sql<Session[]>`
        UPDATE sessions
        SET
@@ -167,10 +266,11 @@ export function registerSessionRoutes(
          system_prompt = COALESCE(${system_prompt ?? null}, system_prompt),
          agent_id = CASE WHEN ${agentIdProvided} THEN ${newAgentId} ELSE agent_id END,
          web_search_enabled = CASE WHEN ${wseProvided} THEN ${newWse} ELSE web_search_enabled END,
+          allowed_read_paths = CASE WHEN ${arpProvided} THEN ${sql.array(newArp, 25)} ELSE allowed_read_paths END,
          updated_at = clock_timestamp()
        WHERE id = ${req.params.id}
        RETURNING id, project_id, name, model, system_prompt, status, created_at, updated_at,
-                  agent_id, web_search_enabled, workspace_panes
+                  agent_id, web_search_enabled, workspace_panes, allowed_read_paths
      `;
      if (rows.length === 0) {
        reply.code(404);
@@ -213,7 +313,7 @@ export function registerSessionRoutes(
            updated_at = clock_timestamp()
        WHERE id = ${req.params.id}
        RETURNING id, project_id, name, model, system_prompt, status, created_at, updated_at,
-                  agent_id, web_search_enabled, workspace_panes
+                  agent_id, web_search_enabled, workspace_panes, allowed_read_paths
      `;
      if (rows.length === 0) {
        reply.code(404);
--- a/apps/server/src/routes/skills.ts
+++ b/apps/server/src/routes/skills.ts
@@ -86,12 +86,12 @@ export function registerSkillsRoutes(

      const result = await sql.begin(async (tx) => {
        const [synthAssistant] = await tx<{ id: string }[]>`
-          INSERT INTO messages (session_id, chat_id, role, content, tool_calls, status, created_at)
-          VALUES (${sessionId}, ${chat.id}, 'assistant', '', ${sql.json(toolCalls as never)}, 'complete', clock_timestamp())
+          INSERT INTO messages (session_id, chat_id, role, content, status, created_at)
+          VALUES (${sessionId}, ${chat.id}, 'assistant', '', 'complete', clock_timestamp())
          RETURNING id
        `;
-        // v1.13.0: dual-write the synthetic assistant message's tool_call.
-        // Single skill_use tool_call, no text content, so one part at seq 0.
+        // v1.13.20: parts-only write. Single skill_use tool_call, no text
+        // content, so one part at seq 0.
        await tx`
          INSERT INTO message_parts (message_id, sequence, kind, payload)
          VALUES (${synthAssistant!.id}, 0, 'tool_call', ${tx.json({
@@ -101,11 +101,11 @@ export function registerSkillsRoutes(
          } as never)})
        `;
        const [toolMsg] = await tx<{ id: string }[]>`
-          INSERT INTO messages (session_id, chat_id, role, content, tool_results, status, created_at)
-          VALUES (${sessionId}, ${chat.id}, 'tool', '', ${sql.json(toolResults as never)}, 'complete', clock_timestamp())
+          INSERT INTO messages (session_id, chat_id, role, content, status, created_at)
+          VALUES (${sessionId}, ${chat.id}, 'tool', '', 'complete', clock_timestamp())
          RETURNING id
        `;
-        // v1.13.0: dual-write the synthetic tool result (the skill body).
+        // v1.13.20: parts-only write of the synthetic tool result (skill body).
        await tx`
          INSERT INTO message_parts (message_id, sequence, kind, payload)
          VALUES (${toolMsg!.id}, 0, 'tool_result', ${tx.json(toolResults as never)})
--- a/apps/server/src/schema.sql
+++ b/apps/server/src/schema.sql
@@ -51,7 +51,7 @@ CREATE TABLE IF NOT EXISTS message_parts (
  kind text NOT NULL,
  payload jsonb NOT NULL,
  created_at timestamptz NOT NULL DEFAULT clock_timestamp(),
-  CONSTRAINT message_parts_kind_chk CHECK (kind IN ('text', 'tool_call', 'tool_result', 'reasoning', 'step_start', 'synthesis')),
+  CONSTRAINT message_parts_kind_chk CHECK (kind IN ('text', 'tool_call', 'tool_result', 'reasoning', 'step_start', 'synthesis', 'html_artifact')),
  CONSTRAINT message_parts_seq_uniq UNIQUE (message_id, sequence)
 );
 CREATE INDEX IF NOT EXISTS message_parts_msg_seq_idx ON message_parts (message_id, sequence);
@@ -79,6 +79,10 @@ CREATE INDEX IF NOT EXISTS message_parts_hidden_idx
 -- 'synthesis'; drop + re-add the constraint with the extended enum. Fresh
 -- installs hit the inline constraint above (already updated) and skip this
 -- block via the pg_constraint guard.
+-- v1.14.x-html-artifact-panes: extend the same constraint with 'html_artifact'.
+-- DROP IF EXISTS + DO $$ pg_constraint $$ guard remains idempotent across
+-- both v1.13.13 and v1.14.x boots; the IN list below is the union of every
+-- kind ever shipped.
 ALTER TABLE message_parts DROP CONSTRAINT IF EXISTS message_parts_kind_chk;
 DO $$
 BEGIN
@@ -87,55 +91,48 @@ BEGIN
  ) THEN
    ALTER TABLE message_parts
      ADD CONSTRAINT message_parts_kind_chk
-      CHECK (kind IN ('text', 'tool_call', 'tool_result', 'reasoning', 'step_start', 'synthesis'));
+      CHECK (kind IN ('text', 'tool_call', 'tool_result', 'reasoning', 'step_start', 'synthesis', 'html_artifact'));
  END IF;
 END $$;

 -- v1.13.1-B: read-path view. Read sites SELECT FROM messages_with_parts
 -- instead of messages so tool_calls / tool_results / reasoning_parts come
-- from the granular message_parts table. The COALESCE means pre-v1.13.0
-- history (no parts rows) still resolves via the legacy JSON columns; the
-- dual-write from v1.13.0 keeps both in sync for all rows written since.
-- Writes continue to target `messages` directly — the view is read-only.
-- Shapes match the in-memory ToolCall / ToolResult types: tool_calls is a
-- jsonb array of {id, name, args}, tool_results is a single jsonb object
-- {tool_call_id, output, truncated, error?}. reasoning_parts is new — only
-- consumed by the inference history fetch (payload.ts) so v1.13.1-C can
-- wire reasoning into the model payload. Not surfaced in external APIs yet.
+-- from the granular message_parts table.
+-- v1.13.20: post column-drop. The legacy COALESCE fallback over
+-- messages.tool_calls / messages.tool_results was removed because those
+-- columns no longer exist on the table (see the ALTER TABLE DROP COLUMN
+-- statements below). Writes continue to target `messages` directly — the
+-- view is read-only. Shapes match the in-memory ToolCall / ToolResult
+-- types: tool_calls is a jsonb array of {id, name, args}, tool_results is
+-- a single jsonb object {tool_call_id, output, truncated, error?}.
+-- reasoning_parts is consumed by the inference history fetch (payload.ts)
+-- for v1.13.1-C reasoning round-tripping. Not surfaced in external APIs.
 CREATE OR REPLACE VIEW messages_with_parts AS
 SELECT
  m.id, m.session_id, m.chat_id, m.role, m.content, m.kind, m.status,
  m.last_seq, m.tokens_used, m.ctx_used, m.ctx_max,
  m.started_at, m.finished_at, m.created_at, m.metadata,
  m.summary, m.tail_start_id, m.compacted_at,
-  -- v1.13.4: prune semantics need to distinguish "no parts row exists"
-  -- (pre-v1.13.0 fallback to legacy column) from "all parts hidden"
-  -- (prune intended — return null/empty so the row drops from the model
-  -- payload). A naive COALESCE would fall back to the legacy column when
-  -- every part is hidden, undoing the prune. CASE on EXISTS(any kind)
-  -- splits the two cases.
-  CASE
-    WHEN EXISTS (SELECT 1 FROM message_parts pp
-                  WHERE pp.message_id = m.id AND pp.kind = 'tool_call')
-    THEN (SELECT jsonb_agg(p.payload ORDER BY p.sequence)
-            FROM message_parts p
-           WHERE p.message_id = m.id AND p.kind = 'tool_call' AND p.hidden_at IS NULL)
-    ELSE m.tool_calls
-  END AS tool_calls,
-  CASE
-    WHEN EXISTS (SELECT 1 FROM message_parts pp
-                  WHERE pp.message_id = m.id AND pp.kind = 'tool_result')
-    THEN (SELECT p.payload
-            FROM message_parts p
-           WHERE p.message_id = m.id AND p.kind = 'tool_result' AND p.hidden_at IS NULL
-           ORDER BY p.sequence LIMIT 1)
-    ELSE m.tool_results
-  END AS tool_results,
+  (SELECT jsonb_agg(p.payload ORDER BY p.sequence)
+     FROM message_parts p
+    WHERE p.message_id = m.id AND p.kind = 'tool_call' AND p.hidden_at IS NULL) AS tool_calls,
+  (SELECT p.payload
+     FROM message_parts p
+    WHERE p.message_id = m.id AND p.kind = 'tool_result' AND p.hidden_at IS NULL
+    ORDER BY p.sequence LIMIT 1) AS tool_results,
  (SELECT jsonb_agg(p.payload ORDER BY p.sequence)
     FROM message_parts p
    WHERE p.message_id = m.id AND p.kind = 'reasoning' AND p.hidden_at IS NULL) AS reasoning_parts
 FROM messages m;

+-- v1.13.20: drop legacy tool_calls/tool_results columns. Reads have routed
+-- through messages_with_parts since v1.13.1-B; dual-writes removed in this
+-- batch. The view above was simplified to remove COALESCE fallbacks before
+-- this drop (Postgres rejects column-drop on view-referenced columns).
+-- Idempotent via IF EXISTS.
+ALTER TABLE messages DROP COLUMN IF EXISTS tool_calls;
+ALTER TABLE messages DROP COLUMN IF EXISTS tool_results;
+
 -- v1.13.10: per-tool token cost rolling window. Derives from
 -- messages_with_parts (the v1.13.1-B view that COALESCEs message_parts over
 -- the legacy JSON column) so this works whether the chat predates v1.13.0
@@ -286,19 +283,6 @@ BEGIN
  END IF;
 END $$;

-- v1.12.1: drop stale inline CHECK constraints that were superseded by the
-- named *_chk variants above. messages_status_check missed 'cancelled' and
-- messages_role_check missed 'system' — both narrower than what's in use.
-DO $$
-BEGIN
-  IF EXISTS (SELECT 1 FROM pg_constraint WHERE conname = 'messages_status_check') THEN
-    ALTER TABLE messages DROP CONSTRAINT messages_status_check;
-  END IF;
-  IF EXISTS (SELECT 1 FROM pg_constraint WHERE conname = 'messages_role_check') THEN
-    ALTER TABLE messages DROP CONSTRAINT messages_role_check;
-  END IF;
-END $$;
-
 -- v1.2-project-ux: projects.status + projects.gitea_remote
 -- KEEP IN SYNC: apps/server/src/types/api.ts PROJECT_STATUSES
 ALTER TABLE projects ADD COLUMN IF NOT EXISTS status TEXT NOT NULL DEFAULT 'open';
@@ -330,6 +314,16 @@ END $$;
 -- agent_id is the slugified agent name. NULL means "use BooCode defaults".
 ALTER TABLE sessions ADD COLUMN IF NOT EXISTS agent_id TEXT;

+-- v1.13.17-cross-repo-reads: session-scoped read grants for paths outside the
+-- session's primary project root. Populated only by the request_read_access
+-- tool's approve branch; revoked via PATCH /api/sessions/:id. Values are
+-- absolute paths to project roots OR repo-shaped dirs under
+-- PROJECT_ROOT_WHITELIST (default /opt). No CHECK constraint — validation
+-- happens at write time in services/grant_resolver.ts. Cleared automatically
+-- when the session row is deleted (no cascade needed; the column goes with it).
+ALTER TABLE sessions
+  ADD COLUMN IF NOT EXISTS allowed_read_paths TEXT[] NOT NULL DEFAULT ARRAY[]::TEXT[];
+
 -- v1.8.2: per-message metadata for sentinels (cap-hit) and structured error
 -- reasons. JSONB so future kinds can extend without further schema churn.
 -- Shape for cap_hit:  { kind: 'cap_hit', used: number, limit: number,
--- a/apps/server/src/services/tests/artifacts.test.ts
+++ b/apps/server/src/services/tests/artifacts.test.ts
@@ -0,0 +1,261 @@
+import { mkdtemp, mkdir, readFile, rm, symlink } from 'node:fs/promises';
+import { tmpdir } from 'node:os';
+import { join } from 'node:path';
+import { afterEach, beforeEach, describe, expect, it } from 'vitest';
+import {
+  decideHtmlArtifactWrite,
+  deriveHtmlSlug,
+  deriveHtmlTitle,
+  deriveMarkdownSlug,
+  detectHtmlArtifact,
+  HTML_ARTIFACT_MAX_BYTES,
+  writeHtmlArtifact,
+  writeMarkdownArtifact,
+} from '../artifacts.js';
+import { PathScopeError } from '../path_guard.js';
+
+describe('deriveMarkdownSlug', () => {
+  it('uses the first # heading when present', () => {
+    expect(deriveMarkdownSlug('# Hello World\n\nbody')).toBe('hello-world');
+  });
+
+  it('falls back to first 6 words', () => {
+    const s = deriveMarkdownSlug('the quick brown fox jumps over the lazy dog');
+    expect(s).toBe('the-quick-brown-fox-jumps-over');
+  });
+
+  it('returns "artifact" for empty input', () => {
+    expect(deriveMarkdownSlug('')).toBe('artifact');
+  });
+
+  it('caps at 60 chars and lowercases', () => {
+    const long = '# ' + 'A'.repeat(200);
+    const s = deriveMarkdownSlug(long);
+    expect(s.length).toBeLessThanOrEqual(60);
+    expect(s).toMatch(/^[a-z0-9-]+$/);
+  });
+
+  it('strips trailing punctuation', () => {
+    expect(deriveMarkdownSlug('# Hello, World!!!')).toBe('hello-world');
+  });
+});
+
+describe('deriveHtmlSlug', () => {
+  it('prefers payload.title when set', () => {
+    expect(
+      deriveHtmlSlug({ html_content: '<html></html>', title: 'My Title' }),
+    ).toBe('my-title');
+  });
+
+  it('falls back to <title> tag', () => {
+    expect(
+      deriveHtmlSlug({
+        html_content: '<html><head><title>Page Title</title></head></html>',
+        title: null,
+      }),
+    ).toBe('page-title');
+  });
+
+  it('falls back to first <h1> when no <title>', () => {
+    expect(
+      deriveHtmlSlug({
+        html_content: '<html><body><h1>Heading One</h1></body></html>',
+        title: null,
+      }),
+    ).toBe('heading-one');
+  });
+
+  it('falls back to inner text words', () => {
+    expect(
+      deriveHtmlSlug({
+        html_content: '<div>one two three four five six seven</div>',
+        title: null,
+      }),
+    ).toBe('one-two-three-four-five-six');
+  });
+});
+
+describe('deriveHtmlTitle', () => {
+  it('returns <title> content', () => {
+    expect(deriveHtmlTitle('<html><head><title>T</title></head></html>')).toBe('T');
+  });
+
+  it('falls back to <h1>', () => {
+    expect(deriveHtmlTitle('<body><h1>H</h1></body>')).toBe('H');
+  });
+
+  it('falls back to first 80 chars of inner text', () => {
+    const html = '<div>' + 'x '.repeat(100) + '</div>';
+    const t = deriveHtmlTitle(html);
+    expect(t).not.toBeNull();
+    expect(t!.length).toBeLessThanOrEqual(80);
+  });
+
+  it('returns null for empty html', () => {
+    expect(deriveHtmlTitle('')).toBeNull();
+  });
+});
+
+describe('detectHtmlArtifact', () => {
+  it('detects <!DOCTYPE html> prefix case-insensitively', () => {
+    const html = '<!doctype HTML><html><body>x</body></html>';
+    expect(detectHtmlArtifact(html)).toBe(html);
+  });
+
+  it('strips leading/trailing whitespace before matching', () => {
+    const html = '\n\n<!DOCTYPE html>\n<html></html>\n';
+    expect(detectHtmlArtifact(html)).toBe(html.trim());
+  });
+
+  it('detects fenced ```html block wrapping entire message', () => {
+    const wrapped = '```html\n<!DOCTYPE html>\n<html></html>\n```';
+    expect(detectHtmlArtifact(wrapped)).toContain('<!DOCTYPE html>');
+  });
+
+  it('rejects plain markdown', () => {
+    expect(detectHtmlArtifact('# heading\n\nsome text')).toBeNull();
+  });
+
+  it('rejects message with prose before the doctype', () => {
+    expect(
+      detectHtmlArtifact('Here you go: <!DOCTYPE html><html></html>'),
+    ).toBeNull();
+  });
+
+  it('rejects empty input', () => {
+    expect(detectHtmlArtifact('')).toBeNull();
+    expect(detectHtmlArtifact('   \n  ')).toBeNull();
+  });
+
+  it('rejects fenced block without doctype/<html>', () => {
+    expect(detectHtmlArtifact('```html\n<div>x</div>\n```')).toBeNull();
+  });
+
+  it('accepts fenced block containing <html> tag (no doctype)', () => {
+    const r = detectHtmlArtifact('```html\n<html><body>x</body></html>\n```');
+    expect(r).toContain('<html>');
+  });
+});
+
+describe('writeMarkdownArtifact / writeHtmlArtifact', () => {
+  let projectRoot: string;
+  beforeEach(async () => {
+    projectRoot = await mkdtemp(join(tmpdir(), 'artifacts-test-'));
+  });
+  afterEach(async () => {
+    await rm(projectRoot, { recursive: true, force: true });
+  });
+
+  it('writes a markdown artifact under .boocode/artifacts/', async () => {
+    const result = await writeMarkdownArtifact(
+      { content: '# Hello\n\nbody' },
+      { projectId: 'pid', projectRoot },
+    );
+    expect(result.path).toMatch(/\.boocode\/artifacts\/hello-\d+\.md$/);
+    expect(result.url).toMatch(/^\/api\/projects\/pid\/artifacts\/hello-\d+\.md$/);
+    const written = await readFile(result.path, 'utf8');
+    expect(written).toBe('# Hello\n\nbody');
+  });
+
+  it('writes an html artifact', async () => {
+    const result = await writeHtmlArtifact(
+      {
+        html_content: '<!DOCTYPE html><html><head><title>X</title></head></html>',
+        char_count: 56,
+        title: 'X',
+      },
+      { projectId: 'pid', projectRoot },
+    );
+    expect(result.path).toMatch(/\.boocode\/artifacts\/x-\d+\.html$/);
+    const written = await readFile(result.path, 'utf8');
+    expect(written).toContain('<!DOCTYPE html>');
+  });
+
+  it('creates the artifacts directory if absent', async () => {
+    // Confirm the writer mkdir-recursive's the artifacts dir on first call.
+    const result = await writeMarkdownArtifact(
+      { content: '# T' },
+      { projectId: 'pid', projectRoot },
+    );
+    expect(result.path).toContain('.boocode/artifacts');
+  });
+});
+
+describe('1MB cap behavior', () => {
+  it('reports the correct byte threshold', () => {
+    expect(HTML_ARTIFACT_MAX_BYTES).toBe(1_048_576);
+  });
+
+  it('exceeds threshold for oversize payload', () => {
+    const oversize = '<!DOCTYPE html>' + 'A'.repeat(HTML_ARTIFACT_MAX_BYTES);
+    expect(Buffer.byteLength(oversize, 'utf8')).toBeGreaterThan(
+      HTML_ARTIFACT_MAX_BYTES,
+    );
+  });
+
+  it('detectHtmlArtifact still returns content above the cap (cap is checked by caller)', () => {
+    // Detection is content-shape; the cap check lives in finalizeCompletion
+    // (error-handler.ts). This test pins that contract: the helper does not
+    // silently drop oversize payloads on the floor.
+    const big = '<!DOCTYPE html>' + 'x'.repeat(2_000_000);
+    expect(detectHtmlArtifact(big)).not.toBeNull();
+  });
+});
+
+describe('decideHtmlArtifactWrite', () => {
+  // Pure helper extracted from finalizeCompletion's cap-skip branch. Pins
+  // the warn-and-skip decision without mocking the full InferenceContext.
+  it('returns write=true for payloads under the cap', () => {
+    const html = '<!DOCTYPE html><html></html>';
+    const decision = decideHtmlArtifactWrite(html);
+    expect(decision.write).toBe(true);
+    expect(decision.byteLen).toBe(Buffer.byteLength(html, 'utf8'));
+  });
+
+  it('returns write=false with cap_exceeded reason for oversize payloads', () => {
+    const big = '<!DOCTYPE html>' + 'x'.repeat(HTML_ARTIFACT_MAX_BYTES);
+    const decision = decideHtmlArtifactWrite(big);
+    expect(decision.write).toBe(false);
+    if (!decision.write) {
+      expect(decision.reason).toBe('cap_exceeded');
+      expect(decision.byteLen).toBeGreaterThan(HTML_ARTIFACT_MAX_BYTES);
+    }
+  });
+
+  it('accepts payload exactly at the cap (boundary)', () => {
+    // byteLen === cap should write; only strictly greater skips.
+    const exact = 'x'.repeat(HTML_ARTIFACT_MAX_BYTES);
+    const decision = decideHtmlArtifactWrite(exact);
+    expect(decision.write).toBe(true);
+    expect(decision.byteLen).toBe(HTML_ARTIFACT_MAX_BYTES);
+  });
+});
+
+describe('symlink escape protection', () => {
+  // Closes the gap where `.boocode/artifacts` is a symlink pointing
+  // outside the project root. The lexical prefix check on the resolved
+  // candidate path passes (it's under projectRoot textually), but the
+  // post-mkdir realpath verification must catch the escape.
+  let projectRoot: string;
+  let outside: string;
+  beforeEach(async () => {
+    projectRoot = await mkdtemp(join(tmpdir(), 'artifacts-symlink-root-'));
+    outside = await mkdtemp(join(tmpdir(), 'artifacts-symlink-outside-'));
+  });
+  afterEach(async () => {
+    await rm(projectRoot, { recursive: true, force: true });
+    await rm(outside, { recursive: true, force: true });
+  });
+
+  it('throws PathScopeError when .boocode/artifacts is a symlink to outside the project', async () => {
+    // Create .boocode dir, then make `artifacts` a symlink pointing outside.
+    await mkdir(join(projectRoot, '.boocode'), { recursive: true });
+    await symlink(outside, join(projectRoot, '.boocode', 'artifacts'));
+    await expect(
+      writeMarkdownArtifact(
+        { content: '# Hello' },
+        { projectId: 'pid', projectRoot },
+      ),
+    ).rejects.toBeInstanceOf(PathScopeError);
+  });
+});
--- a/apps/server/src/services/tests/codecontext_client.test.ts
+++ b/apps/server/src/services/tests/codecontext_client.test.ts
@@ -1,5 +1,5 @@
 import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
-import { mkdir, mkdtemp, rm } from 'node:fs/promises';
+import { mkdir, mkdtemp, rm, symlink, writeFile } from 'node:fs/promises';
 import { join } from 'node:path';
 import { tmpdir } from 'node:os';
 import { callCodecontext } from '../codecontext_client.js';
@@ -203,3 +203,197 @@ describe('callCodecontext — error paths', () => {
    ).rejects.toThrow(/timed out after 30000ms/);
  });
 });
+
+// ---- v1.13.18: file_path resolution tests -----------------------------------
+
+describe('callCodecontext — file_path resolution', () => {
+  // Case 1: relative path resolves to absolute under project root
+  it('resolves a relative file_path to an absolute path inside project root', async () => {
+    // Create a real file so realpath can canonicalise it
+    const fileName = 'src_module.ts';
+    await writeFile(join(projectDir, fileName), '// hello');
+    const fetcher = vi.fn().mockResolvedValue(
+      mockJSONResponse({ result: 'file analysis', error: null }),
+    );
+    await callCodecontext(
+      {
+        toolName: 'get_file_analysis',
+        args: { file_path: fileName },
+        projectPath: projectDir,
+      },
+      fetcher as unknown as typeof fetch,
+    );
+    expect(fetcher).toHaveBeenCalledTimes(1);
+    const body = JSON.parse(fetcher.mock.calls[0]![1]!.body as string);
+    // Should be the resolved absolute path
+    expect(body.file_path).toBe(join(projectDir, fileName));
+  });
+
+  // Case 2: absolute path inside project root → realpathed → forwarded
+  it('passes through an absolute file_path inside project root', async () => {
+    const fileName = 'absolute_target.ts';
+    const absPath = join(projectDir, fileName);
+    await writeFile(absPath, '// absolute');
+    const fetcher = vi.fn().mockResolvedValue(
+      mockJSONResponse({ result: 'analysis', error: null }),
+    );
+    await callCodecontext(
+      {
+        toolName: 'get_file_analysis',
+        args: { file_path: absPath },
+        projectPath: projectDir,
+      },
+      fetcher as unknown as typeof fetch,
+    );
+    const body = JSON.parse(fetcher.mock.calls[0]![1]!.body as string);
+    expect(body.file_path).toBe(absPath);
+  });
+
+  // Case 3: relative escape path → rejected with same error shape as target_dir escape
+  it('rejects a relative file_path that escapes the project root', async () => {
+    const fetcher = vi.fn();
+    await expect(
+      callCodecontext(
+        {
+          toolName: 'get_file_analysis',
+          args: { file_path: '../../etc/passwd' },
+          projectPath: projectDir,
+        },
+        fetcher as unknown as typeof fetch,
+      ),
+    ).rejects.toThrow(/escapes project root/);
+    expect(fetcher).not.toHaveBeenCalled();
+  });
+
+  // Case 4: absolute path outside project root → rejected
+  it('rejects an absolute file_path outside the project root', async () => {
+    const fetcher = vi.fn();
+    await expect(
+      callCodecontext(
+        {
+          toolName: 'get_file_analysis',
+          // /etc/passwd is outside any tmpdir project root
+          args: { file_path: '/etc/passwd' },
+          projectPath: projectDir,
+        },
+        fetcher as unknown as typeof fetch,
+      ),
+    ).rejects.toThrow(/escapes project root/);
+    expect(fetcher).not.toHaveBeenCalled();
+  });
+
+  // Case 5: nonexistent file (ENOENT) → forwarded as un-realpath'd absolute
+  it('forwards a nonexistent file_path as absolute without throwing', async () => {
+    const missingPath = join(projectDir, 'does_not_exist.ts');
+    const fetcher = vi.fn().mockResolvedValue(
+      mockJSONResponse({ result: null, error: 'File not found in graph: ' + missingPath }),
+    );
+    // The resolver should NOT throw; the error comes back from the sidecar
+    await expect(
+      callCodecontext(
+        {
+          toolName: 'get_file_analysis',
+          args: { file_path: 'does_not_exist.ts' },
+          projectPath: projectDir,
+        },
+        fetcher as unknown as typeof fetch,
+      ),
+    ).rejects.toThrow(/File not found in graph/);
+    // Wire was still called — resolver forwarded the path
+    expect(fetcher).toHaveBeenCalledTimes(1);
+    const body = JSON.parse(fetcher.mock.calls[0]![1]!.body as string);
+    // Should receive the absolute (non-realpathed) path
+    expect(body.file_path).toBe(missingPath);
+  });
+
+  // Case 6: empty string → skipped by guard, reaches wire unmodified
+  // Note: Zod .trim().min(1) in get_file_analysis rejects empty before the
+  // shim is reached in production. At the shim layer, the guard
+  // `file_path.trim() !== ''` skips the resolver for empty strings so that
+  // optional-file_path wrappers treat '' as "not provided". This is a
+  // deliberate design; callers that require file_path validate at the Zod layer.
+  it('skips resolver for empty string file_path (treated as not provided)', async () => {
+    const fetcher = vi.fn().mockResolvedValue(
+      mockJSONResponse({ result: 'analysis', error: null }),
+    );
+    // Should succeed — empty string is treated as "no file_path"
+    await callCodecontext(
+      {
+        toolName: 'get_file_analysis',
+        args: { file_path: '' },
+        projectPath: projectDir,
+      },
+      fetcher as unknown as typeof fetch,
+    );
+    expect(fetcher).toHaveBeenCalledTimes(1);
+    const body = JSON.parse(fetcher.mock.calls[0]![1]!.body as string);
+    // Empty string passes through unchanged (resolver not invoked)
+    expect(body.file_path).toBe('');
+  });
+
+  // Case 7: wrapper without file_path (e.g. get_codebase_overview) → resolver not invoked
+  it('does not invoke file_path resolver when file_path is absent from args', async () => {
+    const fetcher = vi.fn().mockResolvedValue(
+      mockJSONResponse({ result: 'overview', error: null }),
+    );
+    await callCodecontext(
+      {
+        toolName: 'get_codebase_overview',
+        args: { include_stats: true },
+        projectPath: projectDir,
+      },
+      fetcher as unknown as typeof fetch,
+    );
+    expect(fetcher).toHaveBeenCalledTimes(1);
+    const body = JSON.parse(fetcher.mock.calls[0]![1]!.body as string);
+    // No file_path in the wire body
+    expect('file_path' in body).toBe(false);
+  });
+
+  // Case 8: absolute path with `..` that resolves outside project root, even
+  // when the literal path is ENOENT. Without resolve() in the absolute branch
+  // the prefix check false-positives because the raw `<projectDir>/../etc/x`
+  // literal starts with `<projectDir>/`.
+  it('rejects absolute file_path with `..` resolving outside project root (ENOENT branch)', async () => {
+    const fetcher = vi.fn();
+    const escapingAbsolute = `${projectDir}/../etc/non_existent_passwd`;
+    await expect(
+      callCodecontext(
+        {
+          toolName: 'get_file_analysis',
+          args: { file_path: escapingAbsolute },
+          projectPath: projectDir,
+        },
+        fetcher as unknown as typeof fetch,
+      ),
+    ).rejects.toThrow(/escapes project root/);
+    expect(fetcher).not.toHaveBeenCalled();
+  });
+
+  // Case 9: in-project symlink targeting outside the project root. This is the
+  // canonical realpath defense — realpath must canonicalise the symlink and
+  // the escape check must reject. Without this test, a symlink-out hole could
+  // regress silently.
+  it('rejects file_path that resolves through a symlink leaving project root', async () => {
+    const outsideDir = await mkdtemp(join(tmpdir(), 'codecontext-outside-'));
+    try {
+      const evilTarget = join(outsideDir, 'secrets.txt');
+      await writeFile(evilTarget, 'top secret');
+      await symlink(evilTarget, join(projectDir, 'evil-link'));
+      const fetcher = vi.fn();
+      await expect(
+        callCodecontext(
+          {
+            toolName: 'get_file_analysis',
+            args: { file_path: 'evil-link' },
+            projectPath: projectDir,
+          },
+          fetcher as unknown as typeof fetch,
+        ),
+      ).rejects.toThrow(/escapes project root/);
+      expect(fetcher).not.toHaveBeenCalled();
+    } finally {
+      await rm(outsideDir, { recursive: true, force: true });
+    }
+  });
+});
--- a/apps/server/src/services/tests/codecontext_tools.test.ts
+++ b/apps/server/src/services/tests/codecontext_tools.test.ts
@@ -70,7 +70,7 @@ describe('codecontext wrappers — toolName + args forwarding', () => {
    const { url, body } = parsePOST(fetcher);
    expect(url).toMatch(/\/v1\/get_file_analysis$/);
    expect(body).toMatchObject({
-      file_path: 'apps/server/src/index.ts',
+      file_path: join(projectDir, 'apps/server/src/index.ts'),
      target_dir: projectDir,
    });
  });
--- a/apps/server/src/services/tests/grant_resolver.test.ts
+++ b/apps/server/src/services/tests/grant_resolver.test.ts
@@ -0,0 +1,199 @@
+// v1.13.17-cross-repo-reads: resolveGrantRoot decision tree.
+//
+// Sam's dispatch note (2026-05-22): "in the project-root resolver ancestor
+// walk, stop the moment parent exits PROJECT_ROOT_WHITELIST or hits
+// filesystem root — check on every iteration, not just final parent.
+// Symlinked input must not be able to escape the whitelist during the
+// walk." The symlink-escape-mid-walk test below pins that invariant —
+// without the per-iteration whitelist check, this case would walk OUTSIDE
+// the whitelist root and return a phantom grant.
+
+import { describe, it, expect, beforeAll, afterAll, vi } from 'vitest';
+import { mkdtemp, rm, mkdir, writeFile, symlink } from 'node:fs/promises';
+import { tmpdir } from 'node:os';
+import { join } from 'node:path';
+import { realpath } from 'node:fs/promises';
+import { resolveGrantRoot } from '../grant_resolver.js';
+import type { Sql } from '../../db.js';
+
+let tmp: string;
+let whitelist: string;
+let project: string;
+let fork: string;
+let outside: string;
+
+// Fake sql tag — returns the projects rows we want without touching a real
+// database. The resolver only ever does a single SELECT, so a single-shot
+// mock that returns the prepared rows on every invocation is enough.
+function makeSql(rows: Array<{ path: string }>): Sql {
+  const tag = ((..._args: unknown[]) => Promise.resolve(rows)) as unknown as Sql;
+  return tag;
+}
+
+beforeAll(async () => {
+  tmp = await realpath(await mkdtemp(join(tmpdir(), 'boocode-gr-')));
+  whitelist = join(tmp, 'whitelist');
+  project = join(whitelist, 'boocode');
+  fork = join(whitelist, 'forks', 'codecontext');
+  outside = join(tmp, 'outside');
+  await mkdir(project, { recursive: true });
+  await mkdir(fork, { recursive: true });
+  await mkdir(outside, { recursive: true });
+  // Mark project as a repo (.git directory).
+  await mkdir(join(project, '.git'));
+  await writeFile(join(project, 'README.md'), 'project readme');
+  // Mark fork as a repo via go.mod (matches the proposal's example).
+  await writeFile(join(fork, 'go.mod'), 'module example.com/foo');
+  await writeFile(join(fork, 'main.go'), 'package main');
+  await writeFile(join(outside, 'secret.txt'), 'forbidden');
+});
+
+afterAll(async () => {
+  await rm(tmp, { recursive: true, force: true });
+});
+
+describe('resolveGrantRoot — happy paths', () => {
+  it('refuses when the requested path is already under projectRoot', async () => {
+    const result = await resolveGrantRoot(makeSql([]), join(project, 'README.md'), project, whitelist);
+    expect(result.ok).toBe(false);
+    if (!result.ok) expect(result.reason).toMatch(/already accessible/);
+  });
+
+  it('returns the project root when the path falls under a registered project', async () => {
+    // Register `fork` as a known project. Resolver should return the project
+    // ancestor (LONGEST match wins) rather than the repo-shape fallback.
+    const result = await resolveGrantRoot(
+      makeSql([{ path: fork }]),
+      join(fork, 'main.go'),
+      project,
+      whitelist,
+    );
+    expect(result.ok).toBe(true);
+    if (result.ok) {
+      expect(result.root).toBe(fork);
+      expect(result.source).toBe('project');
+    }
+  });
+
+  it('falls back to the nearest repo-shaped ancestor when no project matches', async () => {
+    const result = await resolveGrantRoot(
+      makeSql([]),
+      join(fork, 'main.go'),
+      project,
+      whitelist,
+    );
+    expect(result.ok).toBe(true);
+    if (result.ok) {
+      expect(result.root).toBe(fork);
+      expect(result.source).toBe('whitelist');
+    }
+  });
+});
+
+describe('resolveGrantRoot — refusals', () => {
+  it('refuses paths outside PROJECT_ROOT_WHITELIST', async () => {
+    const result = await resolveGrantRoot(
+      makeSql([]),
+      join(outside, 'secret.txt'),
+      project,
+      whitelist,
+    );
+    expect(result.ok).toBe(false);
+    if (!result.ok) expect(result.reason).toMatch(/outside permitted scope/);
+  });
+
+  it('refuses non-absolute paths', async () => {
+    const result = await resolveGrantRoot(makeSql([]), 'relative/path', project, whitelist);
+    expect(result.ok).toBe(false);
+    if (!result.ok) expect(result.reason).toMatch(/absolute/);
+  });
+
+  it('refuses missing paths without prompting', async () => {
+    const result = await resolveGrantRoot(
+      makeSql([]),
+      join(whitelist, 'nope'),
+      project,
+      whitelist,
+    );
+    expect(result.ok).toBe(false);
+    if (!result.ok) expect(result.reason).toMatch(/does not exist/);
+  });
+
+  it('refuses when no repo-shape marker is found before hitting the whitelist root', async () => {
+    // Build a directory tree under the whitelist that has NO repo markers
+    // all the way up to the whitelist root.
+    const plain = join(whitelist, 'plain-dir', 'nested');
+    await mkdir(plain, { recursive: true });
+    await writeFile(join(plain, 'just-a-file.txt'), 'x');
+    const result = await resolveGrantRoot(
+      makeSql([]),
+      join(plain, 'just-a-file.txt'),
+      project,
+      whitelist,
+    );
+    expect(result.ok).toBe(false);
+    if (!result.ok) expect(result.reason).toMatch(/no repo-shaped ancestor/);
+  });
+
+  it('does not grant the whitelist root itself as a fallback', async () => {
+    // Even if .git existed at the whitelist root (it doesn't), we'd refuse.
+    // Easier to assert: a path directly under whitelist with no repo marker.
+    const direct = join(whitelist, 'lone-file.txt');
+    await writeFile(direct, 'x');
+    const result = await resolveGrantRoot(makeSql([]), direct, project, whitelist);
+    expect(result.ok).toBe(false);
+  });
+});
+
+describe('resolveGrantRoot — symlink-escape-mid-walk guard (Sam 2026-05-22)', () => {
+  it('refuses a symlinked input whose realpath sits outside the whitelist', async () => {
+    // The symlink lives nominally inside the whitelist, but its target
+    // (realpath) is outside. The guard's first realpath() call normalizes
+    // and the up-front whitelist check refuses immediately.
+    const link = join(whitelist, 'escape-link');
+    try {
+      await symlink(outside, link);
+      const result = await resolveGrantRoot(
+        makeSql([]),
+        join(link, 'secret.txt'),
+        project,
+        whitelist,
+      );
+      expect(result.ok).toBe(false);
+      if (!result.ok) expect(result.reason).toMatch(/outside permitted scope/);
+    } finally {
+      await rm(link, { force: true });
+    }
+  });
+
+  it('walk loop terminates at the whitelist root, not at filesystem /', async () => {
+    // Construct a deep tree with NO repo markers anywhere. Without a bound,
+    // the walk would chase parents up to "/". The bound flips the loop into
+    // a refusal once the cursor equals the realpath'd whitelist root.
+    const deep = join(whitelist, 'a', 'b', 'c', 'd');
+    await mkdir(deep, { recursive: true });
+    await writeFile(join(deep, 'leaf.txt'), 'x');
+    const result = await resolveGrantRoot(makeSql([]), join(deep, 'leaf.txt'), project, whitelist);
+    expect(result.ok).toBe(false);
+    if (!result.ok) expect(result.reason).toMatch(/no repo-shaped ancestor/);
+  });
+});
+
+describe('resolveGrantRoot — nearest-project disambiguation', () => {
+  it('prefers the longest matching project path over a shorter ancestor', async () => {
+    const outer = whitelist;
+    const inner = fork; // /whitelist/forks/codecontext, deeper than outer
+    const result = await resolveGrantRoot(
+      makeSql([{ path: outer }, { path: inner }]),
+      join(fork, 'main.go'),
+      project,
+      whitelist,
+    );
+    expect(result.ok).toBe(true);
+    if (result.ok) expect(result.root).toBe(inner);
+  });
+});
+
+// Belt-and-suspenders: silence a known dynamic-import warning that vitest
+// occasionally emits on transient fs operations in CI but never in dev.
+vi.spyOn(console, 'warn').mockImplementation(() => {});
--- a/apps/server/src/services/tests/mcp-client.test.ts
+++ b/apps/server/src/services/tests/mcp-client.test.ts
@@ -0,0 +1,169 @@
+/**
+ * v1.15.0-mcp-multi: unit tests for the multi-server MCP client.
+ * Pure unit tests — no live MCP server needed. Tests tool-wrapping,
+ * read-only guard, name prefixing, content extraction, and error handling.
+ * Multi-server routing tested via wrapMcpTool's server-name prefix.
+ */
+import { describe, it, expect } from 'vitest';
+import { wrapMcpTool, extractContent, isToolReadOnly } from '../mcp-client.js';
+
+describe('mcp-client', () => {
+  describe('wrapMcpTool — multi-server prefixing', () => {
+    it('produces a ToolDef with <serverName>_ prefix', () => {
+      const mcpTool = {
+        name: 'resolve-library-id',
+        description: 'Resolve a library identifier',
+        inputSchema: {
+          type: 'object' as const,
+          properties: { query: { type: 'string' } },
+          required: ['query'],
+        },
+      };
+
+      const wrapped = wrapMcpTool('context7', mcpTool);
+
+      expect(wrapped.name).toBe('context7_resolve-library-id');
+      expect(wrapped.description).toBe('Resolve a library identifier');
+      expect(wrapped.jsonSchema.type).toBe('function');
+      expect(wrapped.jsonSchema.function.name).toBe('context7_resolve-library-id');
+      expect(wrapped.jsonSchema.function.parameters).toEqual(mcpTool.inputSchema);
+      expect(typeof wrapped.execute).toBe('function');
+    });
+
+    it('prefixes tools from different servers correctly', () => {
+      const toolA = {
+        name: 'query-docs',
+        description: 'Query docs',
+        inputSchema: { type: 'object' as const, properties: {} },
+      };
+      const toolB = {
+        name: 'overview',
+        description: 'Get overview',
+        inputSchema: { type: 'object' as const, properties: {} },
+      };
+
+      const wrappedA = wrapMcpTool('context7', toolA);
+      const wrappedB = wrapMcpTool('codecontext', toolB);
+
+      expect(wrappedA.name).toBe('context7_query-docs');
+      expect(wrappedB.name).toBe('codecontext_overview');
+    });
+
+    it('multi-server: two servers with 2 tools each produce 4 prefixed tools', () => {
+      const serverATools = [
+        { name: 'query-docs', inputSchema: { type: 'object' as const, properties: {} } },
+        { name: 'resolve-library-id', inputSchema: { type: 'object' as const, properties: {} } },
+      ];
+      const serverBTools = [
+        { name: 'overview', inputSchema: { type: 'object' as const, properties: {} } },
+        { name: 'search', inputSchema: { type: 'object' as const, properties: {} } },
+      ];
+
+      const allWrapped = [
+        ...serverATools.map((t) => wrapMcpTool('context7', t)),
+        ...serverBTools.map((t) => wrapMcpTool('codecontext', t)),
+      ];
+
+      expect(allWrapped).toHaveLength(4);
+      expect(allWrapped.map((t) => t.name)).toEqual([
+        'context7_query-docs',
+        'context7_resolve-library-id',
+        'codecontext_overview',
+        'codecontext_search',
+      ]);
+    });
+
+    it('defaults description to empty string when absent', () => {
+      const mcpTool = {
+        name: 'no-desc',
+        inputSchema: { type: 'object' as const, properties: {} },
+      };
+
+      const wrapped = wrapMcpTool('myserver', mcpTool);
+
+      expect(wrapped.description).toBe('');
+      expect(wrapped.jsonSchema.function.description).toBe('');
+    });
+
+    it('uses passthrough Zod schema (z.record)', () => {
+      const mcpTool = {
+        name: 'test',
+        inputSchema: { type: 'object' as const, properties: {} },
+      };
+
+      const wrapped = wrapMcpTool('s', mcpTool);
+
+      const result = wrapped.inputSchema.safeParse({ foo: 'bar', baz: 123 });
+      expect(result.success).toBe(true);
+    });
+  });
+
+  describe('isToolReadOnly', () => {
+    it('accepts tools with readOnlyHint: true', () => {
+      expect(isToolReadOnly({ readOnlyHint: true })).toBe(true);
+    });
+
+    it('accepts tools with no annotations', () => {
+      expect(isToolReadOnly(undefined)).toBe(true);
+    });
+
+    it('accepts tools with empty annotations', () => {
+      expect(isToolReadOnly({})).toBe(true);
+    });
+
+    it('rejects tools with readOnlyHint: false', () => {
+      expect(isToolReadOnly({ readOnlyHint: false })).toBe(false);
+    });
+
+    it('accepts tools with only destructiveHint set', () => {
+      expect(isToolReadOnly({ destructiveHint: true })).toBe(true);
+    });
+  });
+
+  describe('extractContent', () => {
+    it('extracts single text block', () => {
+      const content = [{ type: 'text', text: 'hello world' }];
+      expect(extractContent(content)).toBe('hello world');
+    });
+
+    it('joins multiple text blocks with newline', () => {
+      const content = [
+        { type: 'text', text: 'line 1' },
+        { type: 'text', text: 'line 2' },
+      ];
+      expect(extractContent(content)).toBe('line 1\nline 2');
+    });
+
+    it('returns "(no output)" for empty content', () => {
+      expect(extractContent([])).toBe('(no output)');
+    });
+
+    it('returns "(no output)" for undefined content', () => {
+      expect(extractContent(undefined)).toBe('(no output)');
+    });
+
+    it('serializes non-text blocks as JSON', () => {
+      const content = [
+        { type: 'resource', uri: 'file:///foo', mimeType: 'text/plain' },
+      ];
+      const result = extractContent(content);
+      expect(result).toContain('"type":"resource"');
+      expect(result).toContain('"uri":"file:///foo"');
+    });
+
+    it('returns error shape when isError is true', () => {
+      const content = [{ type: 'text', text: 'something failed' }];
+      const result = extractContent(content, true);
+      expect(result).toEqual({ error: true, output: 'something failed' });
+    });
+
+    it('returns error shape with joined content on isError', () => {
+      const content = [
+        { type: 'text', text: 'error 1' },
+        { type: 'text', text: 'error 2' },
+      ];
+      const result = extractContent(content, true);
+      expect(result).toEqual({ error: true, output: 'error 1\nerror 2' });
+    });
+  });
+});
--- a/apps/server/src/services/tests/mcp-glob.test.ts
+++ b/apps/server/src/services/tests/mcp-glob.test.ts
@@ -0,0 +1,82 @@
+/**
+ * v1.15.0-mcp-multi: unit tests for matchToolGlob.
+ */
+import { describe, it, expect } from 'vitest';
+import { matchToolGlob } from '../agents.js';
+
+describe('matchToolGlob', () => {
+  it('exact match: "grep" matches "grep"', () => {
+    expect(matchToolGlob('grep', ['grep'])).toBe(true);
+  });
+
+  it('exact match: "grep" does not match "grep2"', () => {
+    expect(matchToolGlob('grep2', ['grep'])).toBe(false);
+  });
+
+  it('exact match: multiple tools', () => {
+    expect(matchToolGlob('grep', ['grep', 'view_file'])).toBe(true);
+    expect(matchToolGlob('view_file', ['grep', 'view_file'])).toBe(true);
+    expect(matchToolGlob('find_files', ['grep', 'view_file'])).toBe(false);
+  });
+
+  it('wildcard: "context7_*" matches "context7_query-docs"', () => {
+    expect(matchToolGlob('context7_query-docs', ['context7_*'])).toBe(true);
+  });
+
+  it('wildcard: "context7_*" matches "context7_resolve-library-id"', () => {
+    expect(matchToolGlob('context7_resolve-library-id', ['context7_*'])).toBe(true);
+  });
+
+  it('wildcard: "context7_*" does not match "codecontext_overview"', () => {
+    expect(matchToolGlob('codecontext_overview', ['context7_*'])).toBe(false);
+  });
+
+  it('wildcard: "view_*" matches "view_file" and "view_truncated_output"', () => {
+    expect(matchToolGlob('view_file', ['view_*'])).toBe(true);
+    expect(matchToolGlob('view_truncated_output', ['view_*'])).toBe(true);
+  });
+
+  it('wildcard: "*" matches everything', () => {
+    expect(matchToolGlob('anything', ['*'])).toBe(true);
+    expect(matchToolGlob('context7_query-docs', ['*'])).toBe(true);
+  });
+
+  it('deny: "!web_*" excludes "web_search"', () => {
+    // With only a deny rule and no prior match, the tool is not matched
+    expect(matchToolGlob('web_search', ['!web_*'])).toBe(false);
+  });
+
+  it('last-match-wins: ["*", "!web_*"] excludes web tools, includes others', () => {
+    expect(matchToolGlob('web_search', ['*', '!web_*'])).toBe(false);
+    expect(matchToolGlob('web_fetch', ['*', '!web_*'])).toBe(false);
+    expect(matchToolGlob('grep', ['*', '!web_*'])).toBe(true);
+    expect(matchToolGlob('context7_query-docs', ['*', '!web_*'])).toBe(true);
+  });
+
+  it('last-match-wins: deny then re-allow', () => {
+    // ["!web_*", "web_search"] — deny all web, then re-allow web_search
+    expect(matchToolGlob('web_search', ['!web_*', 'web_search'])).toBe(true);
+    expect(matchToolGlob('web_fetch', ['!web_*', 'web_fetch'])).toBe(true);
+  });
+
+  it('empty patterns: nothing matches', () => {
+    expect(matchToolGlob('grep', [])).toBe(false);
+    expect(matchToolGlob('anything', [])).toBe(false);
+  });
+
+  it('no-glob fallback: exact-match only, same as pre-v1.15', () => {
+    const patterns = ['grep', 'view_file'];
+    expect(matchToolGlob('grep', patterns)).toBe(true);
+    expect(matchToolGlob('view_file', patterns)).toBe(true);
+    expect(matchToolGlob('find_files', patterns)).toBe(false);
+    expect(matchToolGlob('web_search', patterns)).toBe(false);
+  });
+
+  it('mixed glob and exact patterns', () => {
+    const patterns = ['grep', 'context7_*', '!context7_dangerous'];
+    expect(matchToolGlob('grep', patterns)).toBe(true);
+    expect(matchToolGlob('context7_query-docs', patterns)).toBe(true);
+    expect(matchToolGlob('context7_dangerous', patterns)).toBe(false);
+    expect(matchToolGlob('view_file', patterns)).toBe(false);
+  });
+});
--- a/apps/server/src/services/tests/path_guard.test.ts
+++ b/apps/server/src/services/tests/path_guard.test.ts
@@ -0,0 +1,93 @@
+// v1.13.17-cross-repo-reads: pathGuard now accepts an optional extraRoots
+// list. Validates the primary-root path stays the source of truth and that
+// extra roots are consulted when (and only when) the primary rejects.
+
+import { describe, it, expect, beforeAll, afterAll } from 'vitest';
+import { mkdtemp, rm, mkdir, writeFile, symlink } from 'node:fs/promises';
+import { tmpdir } from 'node:os';
+import { join } from 'node:path';
+import { realpath } from 'node:fs/promises';
+import { pathGuard, PathScopeError } from '../path_guard.js';
+
+let tmp: string;
+let projectRoot: string;
+let altRoot: string;
+let outsideDir: string;
+
+beforeAll(async () => {
+  tmp = await realpath(await mkdtemp(join(tmpdir(), 'boocode-pg-')));
+  projectRoot = join(tmp, 'project');
+  altRoot = join(tmp, 'alt');
+  outsideDir = join(tmp, 'outside');
+  await mkdir(projectRoot, { recursive: true });
+  await mkdir(altRoot, { recursive: true });
+  await mkdir(outsideDir, { recursive: true });
+  await writeFile(join(projectRoot, 'inside.txt'), 'p');
+  await writeFile(join(altRoot, 'cross.txt'), 'a');
+  await writeFile(join(outsideDir, 'forbidden.txt'), 'x');
+});
+
+afterAll(async () => {
+  await rm(tmp, { recursive: true, force: true });
+});
+
+describe('pathGuard (v1.13.17 extraRoots)', () => {
+  it('accepts paths inside the primary projectRoot', async () => {
+    const real = await pathGuard(projectRoot, 'inside.txt');
+    expect(real).toBe(join(projectRoot, 'inside.txt'));
+  });
+
+  it('rejects paths outside the primary root when no extra roots given', async () => {
+    await expect(pathGuard(projectRoot, join(outsideDir, 'forbidden.txt'))).rejects.toBeInstanceOf(
+      PathScopeError,
+    );
+  });
+
+  it('accepts cross-root paths when the matching extra root is provided', async () => {
+    const real = await pathGuard(projectRoot, join(altRoot, 'cross.txt'), [altRoot]);
+    expect(real).toBe(join(altRoot, 'cross.txt'));
+  });
+
+  it('rejects cross-root paths even with extra roots when no root matches', async () => {
+    await expect(
+      pathGuard(projectRoot, join(outsideDir, 'forbidden.txt'), [altRoot]),
+    ).rejects.toBeInstanceOf(PathScopeError);
+  });
+
+  it('ignores empty-string extra roots silently', async () => {
+    const real = await pathGuard(projectRoot, join(altRoot, 'cross.txt'), ['', altRoot]);
+    expect(real).toBe(join(altRoot, 'cross.txt'));
+  });
+
+  it('error message contains the request_read_access hint when scope rejects', async () => {
+    try {
+      await pathGuard(projectRoot, join(outsideDir, 'forbidden.txt'));
+      throw new Error('should have thrown');
+    } catch (err) {
+      expect(err).toBeInstanceOf(PathScopeError);
+      expect((err as Error).message).toContain('request_read_access');
+    }
+  });
+
+  it('still resolves symlinks before the scope check', async () => {
+    const linkPath = join(projectRoot, 'link-to-outside');
+    await symlink(join(outsideDir, 'forbidden.txt'), linkPath);
+    // Symlink target escapes both primary and the single extra root, so
+    // even though the surface path "looks" inside projectRoot, the real
+    // path resolves outside and the guard rejects.
+    await expect(pathGuard(projectRoot, linkPath, [altRoot])).rejects.toBeInstanceOf(
+      PathScopeError,
+    );
+    // But adding outsideDir as an extra root accepts (realpath inside it).
+    const real = await pathGuard(projectRoot, linkPath, [altRoot, outsideDir]);
+    expect(real).toBe(join(outsideDir, 'forbidden.txt'));
+  });
+
+  it('tries extra roots in order until one accepts', async () => {
+    const real = await pathGuard(projectRoot, join(altRoot, 'cross.txt'), [
+      outsideDir, // rejects
+      altRoot,    // accepts
+    ]);
+    expect(real).toBe(join(altRoot, 'cross.txt'));
+  });
+});
--- a/apps/server/src/services/tests/tool_cost_stats.test.ts
+++ b/apps/server/src/services/tests/tool_cost_stats.test.ts
@@ -78,16 +78,18 @@ describeFn('tool_cost_stats view (v1.13.10)', () => {
      args: {},
    }));
    const created = opts.createdAt ?? new Date();
+    // v1.13.20: parts-only. messages.tool_calls column was dropped; the
+    // tool_cost_stats view reads through messages_with_parts which derives
+    // tool_calls from message_parts rows.
    const rows = await sql<{ id: string }[]>`
      INSERT INTO messages (
        session_id, chat_id, role, content, kind, status,
-        tool_calls, tokens_used, ctx_used,
+        tokens_used, ctx_used,
        metadata, created_at
      )
      VALUES (
        ${sessionId}, ${chatId}, 'assistant', '', 'message',
        ${opts.status ?? 'complete'},
-        ${sql.json(toolCalls as never)},
        ${opts.tokensUsed},
        ${opts.ctxUsed},
        ${opts.metadata ? sql.json(opts.metadata as never) : null},
@@ -95,7 +97,14 @@ describeFn('tool_cost_stats view (v1.13.10)', () => {
      )
      RETURNING id
    `;
-    return rows[0]!.id;
+    const messageId = rows[0]!.id;
+    for (let i = 0; i < toolCalls.length; i++) {
+      await sql`
+        INSERT INTO message_parts (message_id, sequence, kind, payload)
+        VALUES (${messageId}, ${i}, 'tool_call', ${sql.json(toolCalls[i] as never)})
+      `;
+    }
+    return messageId;
  }

  it('returns empty when no tool calls exist for a tool name', async () => {
@@ -197,18 +206,17 @@ describeFn('tool_cost_stats view (v1.13.10)', () => {

  it('reads tool_calls via messages_with_parts (parts-authoritative)', async () => {
    const t = tname('parts');
-    // Insert an assistant row with messages.tool_calls=NULL but a
-    // message_parts row carrying the tool_call. The view reads via
-    // messages_with_parts, which COALESCEs the parts table over the legacy
-    // column — so this row should still aggregate.
+    // v1.13.20: post-column-drop the only source for tool_calls is
+    // message_parts. This test asserts the same path the view always took
+    // (parts-derived), now that the legacy column COALESCE fallback is gone.
    const rows = await sql<{ id: string }[]>`
      INSERT INTO messages (
        session_id, chat_id, role, content, kind, status,
-        tool_calls, tokens_used, ctx_used
+        tokens_used, ctx_used
      )
      VALUES (
        ${sessionId}, ${chatId}, 'assistant', '', 'message', 'complete',
-        NULL, 200, 5000
+        200, 5000
      )
      RETURNING id
    `;
--- a/apps/server/src/services/agents.ts
+++ b/apps/server/src/services/agents.ts
@@ -16,10 +16,62 @@ const CACHE_TTL_MS = 60_000;
 // hand-maintained list drifted (web_search/web_fetch from v1.11.8 + the 8
 // codecontext tools were missing), silently filtering valid tool names out
 // of agents that opted in. Single source of truth is tools.ts now.
-const ALL_TOOL_NAMES: readonly string[] = ALL_TOOLS.map((t) => t.name);
-const DEFAULT_TOOLS: string[] = [...ALL_TOOL_NAMES];
+let ALL_TOOL_NAMES: readonly string[] = ALL_TOOLS.map((t) => t.name);
+let DEFAULT_TOOLS: string[] = [...ALL_TOOL_NAMES];
+
+export function refreshToolNames(): void {
+  ALL_TOOL_NAMES = ALL_TOOLS.map((t) => t.name);
+  DEFAULT_TOOLS = [...ALL_TOOL_NAMES];
+}
 const DEFAULT_TEMPERATURE = 0.7;

+// ---- Tool glob matching (v1.15.0-mcp-multi) --------------------------------
+
+/**
+ * Simple glob match for tool names. Supports `*` as a wildcard for any
+ * characters. No `?` or `**` — tool names are flat (no path separators).
+ */
+function simpleGlobMatch(str: string, pattern: string): boolean {
+  if (pattern === '*') return true;
+  if (!pattern.includes('*')) return str === pattern;
+  // Escape regex metacharacters, then replace escaped \* with .*
+  const regex = new RegExp(
+    '^' + pattern.replace(/[.+?^${}()|[\]\\]/g, '\\$&').replace(/\*/g, '.*') + '$',
+  );
+  return regex.test(str);
+}
+
+/**
+ * Check if a tool name matches a set of glob patterns. Last-match-wins.
+ * Patterns starting with `!` are deny rules.
+ *
+ * Examples:
+ * - `["grep", "view_file"]` — exact-match whitelist (same as pre-v1.15)
+ * - `["context7_*"]` — all tools from the context7 MCP server
+ * - `["*", "!web_*"]` — all tools except web tools
+ * - `[]` — nothing matches (agent gets no tools)
+ */
+export function matchToolGlob(toolName: string, patterns: string[]): boolean {
+  let matched = false;
+  for (const pattern of patterns) {
+    const deny = pattern.startsWith('!');
+    const glob = deny ? pattern.slice(1) : pattern;
+    if (simpleGlobMatch(toolName, glob)) {
+      matched = !deny;
+    }
+  }
+  return matched;
+}
+
+/**
+ * Returns true if a tools: entry is a glob pattern (contains * or starts
+ * with !). Glob patterns can't be validated against the current tool list
+ * since MCP tools are discovered at runtime.
+ */
+function isGlobPattern(entry: string): boolean {
+  return entry.includes('*') || entry.startsWith('!');
+}
+
 export function slugify(name: string): string {
  return name
    .toLowerCase()
@@ -37,6 +89,10 @@ interface ParsedFrontmatter {
  // v1.8.2: optional per-agent tool-loop budget. Absent → inference resolves
  // from the agent's toolset at runtime.
  max_tool_calls?: number;
+  // v1.14.0: optional per-agent step cap. Absent → bounded only by MAX_STEPS
+  // (200) in the outer loop. Integer ≥ 0; steps: 0 means "no tool calls
+  // allowed" — the model responds text-only.
+  steps?: number;
 }

 function stripQuotes(s: string): string {
@@ -112,6 +168,21 @@ function parseFrontmatter(yaml: string): { data: ParsedFrontmatter; errors: stri
      } else {
        errors.push(`max_tool_calls must be an integer 1-100 (got "${valueRaw}")`);
      }
+    } else if (key === 'steps') {
+      // v1.14.0: per-agent step cap for the outer inference loop. Integer ≥ 0.
+      // steps: 0 means "no tool calls allowed" — model responds text-only.
+      // Non-integer or negative values are warned and ignored (falls back to
+      // MAX_STEPS ceiling), matching the max_tool_calls pattern above.
+      const n = Number(valueRaw);
+      if (Number.isInteger(n) && n >= 0) {
+        data.steps = n;
+      } else if (Number.isInteger(n)) {
+        console.warn(
+          `agents: steps ${n} is negative, ignoring (falling back to default)`,
+        );
+      } else {
+        errors.push(`steps must be a non-negative integer (got "${valueRaw}")`);
+      }
    }
    // Unknown keys silently ignored — forward-compat.
  }
@@ -188,10 +259,14 @@ function parseAgentSection(section: RawSection): Omit<Agent, 'source'> {

  // v1.13.15-tools: intersect with BOOCODE_TOOLS tier (ceiling, not expansion).
  // Unset → resolveToolTier returns ALL tool names → no narrowing.
+  // v1.15.0-mcp-multi: glob patterns (entries containing * or starting with !)
+  // pass through unvalidated — MCP tools are discovered at runtime and can't
+  // be checked against ALL_TOOL_NAMES at parse time.
  const tierAllowed = new Set(resolveToolTier(process.env.BOOCODE_TOOLS));
  const filteredTools = Array.isArray(fm.tools)
    ? fm.tools.filter((t): t is string =>
-        (ALL_TOOL_NAMES as readonly string[]).includes(t) && tierAllowed.has(t),
+        isGlobPattern(t) ||
+        ((ALL_TOOL_NAMES as readonly string[]).includes(t) && tierAllowed.has(t)),
      )
    : DEFAULT_TOOLS.filter((t) => tierAllowed.has(t));

@@ -204,6 +279,7 @@ function parseAgentSection(section: RawSection): Omit<Agent, 'source'> {
    tools: filteredTools,
    model: typeof fm.model === 'string' && fm.model.length > 0 ? fm.model : null,
    max_tool_calls: typeof fm.max_tool_calls === 'number' ? fm.max_tool_calls : null,
+    steps: typeof fm.steps === 'number' ? fm.steps : null,
  };
 }

--- a/apps/server/src/services/artifacts.ts
+++ b/apps/server/src/services/artifacts.ts
@@ -0,0 +1,255 @@
+// v1.14.x-html-artifact-panes: artifact writer + slug derivation.
+//
+// Writes Markdown and HTML artifacts to `<projectRoot>/.boocode/artifacts/`
+// as plain files. Returns `{path, url}` where:
+//   - path is the absolute on-disk path
+//   - url is a project-scoped REST URL pointing at the GET download route
+//     registered in routes/artifacts.ts. The route streams the file with
+//     Content-Disposition: attachment.
+//
+// Path safety: we do NOT use path_guard.ts (it realpaths and throws ENOENT
+// for files that don't exist yet, which artifact creation requires).
+// Instead we mirror the v1.13.18 codecontext_client.ts pattern: resolve
+// the candidate path against the realpath'd projectRoot, then verify the
+// result starts with projectRoot + sep (or equals projectRoot).
+
+import { mkdir, realpath, writeFile } from 'node:fs/promises';
+import { resolve, sep } from 'node:path';
+import { PathScopeError } from './path_guard.js';
+import type { Message } from '../types/api.js';
+
+export interface HtmlArtifactPayload {
+  html_content: string;
+  char_count: number;
+  title: string | null;
+}
+
+export interface ArtifactWriteResult {
+  path: string;
+  url: string;
+}
+
+const ARTIFACT_SUBDIR = '.boocode/artifacts';
+
+// ---- slug helpers ----
+
+// Lowercase, replace non-alnum runs with '-', trim leading/trailing '-',
+// collapse repeated '-', cap at 60 chars. Empty → 'artifact'.
+function slugify(input: string): string {
+  const cleaned = input
+    .toLowerCase()
+    .replace(/[^a-z0-9]+/g, '-')
+    .replace(/^-+|-+$/g, '')
+    .replace(/-{2,}/g, '-')
+    .slice(0, 60)
+    .replace(/^-+|-+$/g, '');
+  return cleaned || 'artifact';
+}
+
+function firstHeading(md: string): string | null {
+  // Match the first `# ` ATX heading at the start of a line.
+  const m = md.match(/^[ \t]*#[ \t]+(.+?)\s*$/m);
+  if (!m) return null;
+  const text = m[1]?.trim() ?? '';
+  return text.length > 0 ? text : null;
+}
+
+function firstNWords(s: string, n: number): string {
+  const words = s.trim().split(/\s+/).filter(Boolean).slice(0, n);
+  return words.join(' ');
+}
+
+export function deriveMarkdownSlug(messageContent: string): string {
+  const heading = firstHeading(messageContent);
+  if (heading) return slugify(heading);
+  const sixWords = firstNWords(messageContent, 6);
+  return slugify(sixWords);
+}
+
+// Strip HTML tags for inner-text extraction. Crude but sufficient for slug
+// derivation — we're not rendering, just finding readable words.
+function stripTags(html: string): string {
+  return html
+    .replace(/<script\b[^<]*(?:(?!<\/script>)<[^<]*)*<\/script>/gi, ' ')
+    .replace(/<style\b[^<]*(?:(?!<\/style>)<[^<]*)*<\/style>/gi, ' ')
+    .replace(/<[^>]+>/g, ' ')
+    .replace(/\s+/g, ' ')
+    .trim();
+}
+
+function extractTitleTag(html: string): string | null {
+  const m = html.match(/<title[^>]*>([\s\S]*?)<\/title>/i);
+  if (!m) return null;
+  const text = stripTags(m[1] ?? '').trim();
+  return text.length > 0 ? text : null;
+}
+
+function extractH1(html: string): string | null {
+  const m = html.match(/<h1[^>]*>([\s\S]*?)<\/h1>/i);
+  if (!m) return null;
+  const text = stripTags(m[1] ?? '').trim();
+  return text.length > 0 ? text : null;
+}
+
+export function deriveHtmlSlug(payload: {
+  html_content: string;
+  title: string | null;
+}): string {
+  if (payload.title && payload.title.trim().length > 0) {
+    return slugify(payload.title);
+  }
+  const title = extractTitleTag(payload.html_content);
+  if (title) return slugify(title);
+  const h1 = extractH1(payload.html_content);
+  if (h1) return slugify(h1);
+  const inner = stripTags(payload.html_content);
+  return slugify(firstNWords(inner, 6));
+}
+
+// Derive title for the html_artifact part payload: <title> → first <h1> →
+// first 80 chars of inner text. Returns null if nothing useful is found.
+export function deriveHtmlTitle(html: string): string | null {
+  const t = extractTitleTag(html);
+  if (t) return t;
+  const h1 = extractH1(html);
+  if (h1) return h1;
+  const inner = stripTags(html);
+  if (inner.length === 0) return null;
+  return inner.slice(0, 80);
+}
+
+// ---- HTML detection (B4) ----
+
+// Returns the inner HTML content if `text` is a recognised HTML artifact:
+//   - starts with <!DOCTYPE html> (case-insensitive, whitespace-trimmed), OR
+//   - wrapped entirely in a fenced ```html ... ``` block.
+// Returns null if neither matches.
+export function detectHtmlArtifact(text: string): string | null {
+  const trimmed = text.trim();
+  if (trimmed.length === 0) return null;
+  if (/^<!doctype\s+html/i.test(trimmed)) {
+    return trimmed;
+  }
+  // Fenced ```html block consuming the entire (trimmed) message. Allow an
+  // optional trailing newline before the closing fence.
+  const fence = trimmed.match(/^```html\s*\n([\s\S]*?)\n?```\s*$/i);
+  if (fence) {
+    const inner = fence[1] ?? '';
+    if (/^\s*<!doctype\s+html/i.test(inner) || /<html[\s>]/i.test(inner)) {
+      return inner.trim();
+    }
+  }
+  return null;
+}
+
+// ---- path resolution ----
+
+// Resolve `<projectRoot>/.boocode/artifacts/<filename>` and verify the
+// result stays under projectRoot. Mirrors the v1.13.18 codecontext_client.ts
+// approach: realpath projectRoot first, then prefix-check the candidate.
+// Throws on escape.
+async function resolveArtifactPath(
+  projectRoot: string,
+  filename: string,
+): Promise<{ resolvedRoot: string; artifactsDir: string; absPath: string }> {
+  const resolvedRoot = await realpath(projectRoot);
+  const artifactsDir = resolve(resolvedRoot, ARTIFACT_SUBDIR);
+  const absPath = resolve(artifactsDir, filename);
+  // Lexical prefix check on the resolved candidates. (The `!== resolvedRoot`
+  // branch was dead — ARTIFACT_SUBDIR is non-empty so artifactsDir always
+  // differs from resolvedRoot.)
+  if (!artifactsDir.startsWith(resolvedRoot + sep)) {
+    throw new PathScopeError(
+      `artifacts dir escapes project root: ${artifactsDir}`,
+    );
+  }
+  if (!absPath.startsWith(artifactsDir + sep)) {
+    throw new PathScopeError(
+      `artifact filename escapes artifacts dir: ${filename}`,
+    );
+  }
+  return { resolvedRoot, artifactsDir, absPath };
+}
+
+// After mkdir, realpath the artifacts dir and re-verify it stays under
+// resolvedRoot. Closes the symlink-escape gap: if `.boocode/artifacts` (or
+// any ancestor below resolvedRoot) is a symlink pointing outside the
+// project, the lexical check in resolveArtifactPath passes but the actual
+// write lands outside the sandbox. Throws PathScopeError on escape.
+async function assertArtifactsDirSafe(
+  artifactsDir: string,
+  resolvedRoot: string,
+): Promise<void> {
+  const realDir = await realpath(artifactsDir);
+  if (realDir !== resolvedRoot && !realDir.startsWith(resolvedRoot + sep)) {
+    throw new PathScopeError(
+      `artifacts dir resolves outside project root: ${realDir}`,
+    );
+  }
+}
+
+// Pure decision helper for whether finalizeCompletion should write the
+// `html_artifact` part. Exported for unit testing the cap-skip branch.
+// Returns `{write: true, byteLen}` when the payload is under the cap, or
+// `{write: false, byteLen, reason: 'cap_exceeded'}` when oversize.
+export type HtmlArtifactDecision =
+  | { write: true; byteLen: number }
+  | { write: false; byteLen: number; reason: 'cap_exceeded' };
+
+export function decideHtmlArtifactWrite(
+  htmlContent: string,
+): HtmlArtifactDecision {
+  const byteLen = Buffer.byteLength(htmlContent, 'utf8');
+  if (byteLen > HTML_ARTIFACT_MAX_BYTES) {
+    return { write: false, byteLen, reason: 'cap_exceeded' };
+  }
+  return { write: true, byteLen };
+}
+
+function buildUrl(projectId: string, filename: string): string {
+  return `/api/projects/${projectId}/artifacts/${encodeURIComponent(filename)}`;
+}
+
+export interface WriteContext {
+  projectId: string;
+  projectRoot: string;
+}
+
+export async function writeMarkdownArtifact(
+  message: Pick<Message, 'content'>,
+  ctx: WriteContext,
+): Promise<ArtifactWriteResult> {
+  const slug = deriveMarkdownSlug(message.content);
+  const filename = `${slug}-${Date.now()}.md`;
+  const { resolvedRoot, artifactsDir, absPath } = await resolveArtifactPath(
+    ctx.projectRoot,
+    filename,
+  );
+  await mkdir(artifactsDir, { recursive: true });
+  await assertArtifactsDirSafe(artifactsDir, resolvedRoot);
+  await writeFile(absPath, message.content, 'utf8');
+  return { path: absPath, url: buildUrl(ctx.projectId, filename) };
+}
+
+export async function writeHtmlArtifact(
+  payload: HtmlArtifactPayload,
+  ctx: WriteContext,
+): Promise<ArtifactWriteResult> {
+  const slug = deriveHtmlSlug(payload);
+  const filename = `${slug}-${Date.now()}.html`;
+  const { resolvedRoot, artifactsDir, absPath } = await resolveArtifactPath(
+    ctx.projectRoot,
+    filename,
+  );
+  await mkdir(artifactsDir, { recursive: true });
+  await assertArtifactsDirSafe(artifactsDir, resolvedRoot);
+  await writeFile(absPath, payload.html_content, 'utf8');
+  return { path: absPath, url: buildUrl(ctx.projectId, filename) };
+}
+
+// 1MB cap on HTML artifacts (proposal S6). Larger payloads are not written
+// to the `html_artifact` part — the assistant text lands as plain content
+// and a warning is logged. Streaming abort was considered but the graceful
+// "no artifact, plain text falls back" path is simpler and lossless from
+// the user's perspective.
+export const HTML_ARTIFACT_MAX_BYTES = 1_048_576;
--- a/apps/server/src/services/auto_name.ts
+++ b/apps/server/src/services/auto_name.ts
@@ -1,7 +1,7 @@
 import type { InferenceContext } from './inference/index.js';

 const NAMING_SYSTEM_PROMPT =
-  'You name chat sessions. Reply directly with no thinking, reasoning, or explanation. Output ONLY the title, 4 words max, no quotes, no punctuation, no prefix like "Title:".';
+  'You name chat sessions based on what the assistant did. Summarize the topic or outcome — do NOT copy the first few words verbatim. Reply directly with no thinking, reasoning, or explanation. Output ONLY the title, 4 words max, no quotes, no punctuation, no prefix like "Title:".';

 const MAX_TITLE_CHARS = 60;

@@ -67,15 +67,10 @@ export async function maybeAutoNameChat(
  const sessionRows = await ctx.sql<{ model: string }[]>`
    SELECT model FROM sessions WHERE id = ${sessionId}
  `;
-  const model = sessionRows[0]?.model;
+  // v2.0.5: prefer FAST_MODEL for cheap LLM calls (titles, summaries).
+  const model = ctx.config.FAST_MODEL ?? sessionRows[0]?.model;
  if (!model) return;

-  const userMsg = await ctx.sql<{ content: string }[]>`
-    SELECT content FROM messages
-    WHERE chat_id = ${chatId} AND role = 'user'
-    ORDER BY created_at ASC
-    LIMIT 1
-  `;
  const assistantMsg = await ctx.sql<{ content: string }[]>`
    SELECT content FROM messages
    WHERE chat_id = ${chatId}
@@ -85,9 +80,8 @@ export async function maybeAutoNameChat(
    ORDER BY created_at ASC
    LIMIT 1
  `;
-  if (!userMsg[0] || !assistantMsg[0]) return;
+  if (!assistantMsg[0]) return;

-  const userText = userMsg[0].content.slice(0, 2000);
  const assistantText = assistantMsg[0].content.slice(0, 2000);

  const body = {
@@ -96,7 +90,7 @@ export async function maybeAutoNameChat(
      { role: 'system', content: NAMING_SYSTEM_PROMPT },
      {
        role: 'user',
-        content: `First user message: ${userText}\nFirst assistant reply: ${assistantText}`,
+        content: assistantText,
      },
    ],
    max_tokens: 30,
--- a/apps/server/src/services/codecontext_client.ts
+++ b/apps/server/src/services/codecontext_client.ts
@@ -17,7 +17,7 @@
 //      which we re-surface with a hint to add the file to .codecontextignore.

 import { access, copyFile, realpath } from 'node:fs/promises';
-import { join } from 'node:path';
+import { isAbsolute, join, resolve, sep } from 'node:path';
 import { truncateIfNeeded } from './truncate.js';

 // v1.13.12 fix: codecontext crashes on empty source files (upstream issue #37)
@@ -51,6 +51,45 @@ async function ensureIgnoreFile(projectRoot: string): Promise<void> {
  }
 }

+// v1.13.18: resolve a `file_path` arg to an absolute path anchored within
+// the (already realpath'd) projectRoot. Contract:
+//   - empty/whitespace-only → INVALID_FILE_PATH error
+//   - relative path → resolve(projectRoot, rawPath) (normalises dot-segments)
+//   - absolute path → resolve(rawPath) (also normalises — e.g. /root/../etc
+//     becomes /etc so the prefix-check below rejects it even in the ENOENT
+//     fallthrough where realpath couldn't canonicalise)
+//   - try realpath; on ENOENT fall through with the (normalised) absolute
+//     (the sidecar issues its own "File not found in graph" that the model
+//     can self-correct on; re-implementing the check here would diverge)
+//   - if the final path doesn't sit inside projectRoot → escape error
+//     (same shape as target_dir escape, only the field name differs)
+async function resolveProjectPath(
+  projectRoot: string,
+  rawPath: string,
+): Promise<string> {
+  if (rawPath.trim() === '') {
+    throw new Error('INVALID_FILE_PATH: file_path must not be empty');
+  }
+  const candidate = isAbsolute(rawPath) ? resolve(rawPath) : resolve(projectRoot, rawPath);
+  let resolved: string;
+  try {
+    resolved = await realpath(candidate);
+  } catch (err: unknown) {
+    if ((err as NodeJS.ErrnoException).code === 'ENOENT') {
+      // File doesn't exist yet (or was deleted). Forward the absolute path;
+      // codecontext will return "File not found in graph" which the model
+      // can self-correct on.
+      resolved = candidate;
+    } else {
+      throw err;
+    }
+  }
+  if (resolved !== projectRoot && !resolved.startsWith(projectRoot + sep)) {
+    throw new Error(`file_path ${rawPath} escapes project root ${projectRoot}`);
+  }
+  return resolved;
+}
+
 export interface CodecontextRequest {
  toolName: string;
  args: Record<string, unknown>;
@@ -96,7 +135,14 @@ export async function callCodecontext(

  // Step 2: re-build args with the resolved target_dir so codecontext sees
  // the real absolute path, not a symlink or relative form.
-  const argsToSend = { ...req.args, target_dir: resolvedTarget };
+  // v1.13.18: also resolve file_path when present — the sidecar index is keyed
+  // on absolute paths, so a relative path from the model yields "File not found
+  // in graph". Same escape check as target_dir; ENOENT falls through so the
+  // sidecar produces the canonical "File not found in graph" the model can fix.
+  const argsToSend: Record<string, unknown> = { ...req.args, target_dir: resolvedTarget };
+  if (typeof req.args['file_path'] === 'string' && req.args['file_path'].trim() !== '') {
+    argsToSend['file_path'] = await resolveProjectPath(resolvedProject, req.args['file_path']);
+  }

  // Step 3: POST with a hard timeout. AbortController + setTimeout pattern
  // matches web_fetch.ts; nothing fancier needed.
--- a/apps/server/src/services/file_ops.ts
+++ b/apps/server/src/services/file_ops.ts
@@ -47,8 +47,12 @@ export interface FindFilesResult {
  truncated: boolean;
 }

-export async function listDir(projectRoot: string, relPath: string): Promise<ListDirResult> {
-  const real = await pathGuard(projectRoot, relPath);
+export async function listDir(
+  projectRoot: string,
+  relPath: string,
+  opts?: { extra_roots?: readonly string[] },
+): Promise<ListDirResult> {
+  const real = await pathGuard(projectRoot, relPath, opts?.extra_roots);
  const s = await stat(real);
  if (!s.isDirectory()) {
    throw new PathScopeError(`not a directory: ${relPath}`);
@@ -82,8 +86,12 @@ export async function listDir(projectRoot: string, relPath: string): Promise<Lis
  };
 }

-export async function viewFile(projectRoot: string, relPath: string): Promise<ViewFileResult> {
-  const real = await pathGuard(projectRoot, relPath);
+export async function viewFile(
+  projectRoot: string,
+  relPath: string,
+  opts?: { extra_roots?: readonly string[] },
+): Promise<ViewFileResult> {
+  const real = await pathGuard(projectRoot, relPath, opts?.extra_roots);
  const s = await stat(real);
  if (!s.isFile()) {
    throw new PathScopeError(`not a file: ${relPath}`);
@@ -119,10 +127,10 @@ interface RipgrepMatch {
 export async function grep(
  projectRoot: string,
  pattern: string,
-  opts?: { path?: string; max_matches?: number; case_sensitive?: boolean; hidden?: boolean }
+  opts?: { path?: string; max_matches?: number; case_sensitive?: boolean; hidden?: boolean; extra_roots?: readonly string[] }
 ): Promise<GrepResult> {
  const targetPath = opts?.path ?? projectRoot;
-  const target = await pathGuard(projectRoot, targetPath);
+  const target = await pathGuard(projectRoot, targetPath, opts?.extra_roots);
  const limit = Math.min(
    Math.max(opts?.max_matches ?? DEFAULT_GREP_RESULTS, 1),
    MAX_GREP_RESULTS
@@ -192,14 +200,14 @@ export async function grep(
 export async function findFiles(
  projectRoot: string,
  pattern?: string,
-  opts?: { type?: 'file' | 'dir'; max_results?: number; path?: string }
+  opts?: { type?: 'file' | 'dir'; max_results?: number; path?: string; extra_roots?: readonly string[] }
 ): Promise<FindFilesResult> {
  const limit = Math.min(
    Math.max(opts?.max_results ?? DEFAULT_FIND_RESULTS, 1),
    MAX_FIND_RESULTS
  );
  const target = opts?.path != null
-    ? await pathGuard(projectRoot, opts.path)
+    ? await pathGuard(projectRoot, opts.path, opts?.extra_roots)
    : projectRoot;
  const args = ['--files'];
  if (pattern) args.push('--glob', pattern);
--- a/apps/server/src/services/grant_resolver.ts
+++ b/apps/server/src/services/grant_resolver.ts
@@ -0,0 +1,161 @@
+// v1.13.17-cross-repo-reads: derives the grant root for a path the user is
+// being asked to approve cross-repo read access to.
+//
+// Per design decision D1: grant unit = nearest registered project root,
+// then nearest path-whitelist ancestor that looks like a repo root, then
+// refuse. Granting the literal file path is too narrow (next file in the
+// same repo re-prompts). Granting an arbitrary parent dir over-scopes.
+//
+// The resolver runs in two contexts:
+//   1. request_read_access.execute  — pre-prompt validation (cheap; bails
+//      early if the path can't plausibly be granted so the user is never
+//      asked about /etc/passwd)
+//   2. POST /api/chats/:id/grant_read_access — at decision time, re-derives
+//      the root and persists it on sessions.allowed_read_paths
+//
+// Sam (2026-05-22 dispatch confirmation): "in the project-root resolver
+// ancestor walk, stop the moment parent exits PROJECT_ROOT_WHITELIST or hits
+// filesystem root — check on every iteration, not just final parent.
+// Symlinked input must not be able to escape the whitelist during the
+// walk." Hence the loop here checks both the walk bound AND the still-
+// inside-whitelist invariant every step.
+
+import { access, realpath } from 'node:fs/promises';
+import { constants } from 'node:fs';
+import { dirname, isAbsolute, sep } from 'node:path';
+import type { Sql } from '../db.js';
+
+// Files whose presence in a directory marks it as a repo root for grant
+// purposes. Kept narrow on purpose; broader heuristics (e.g. ".project",
+// "pyproject.toml") can be added with measured intent. Each entry is a
+// literal basename — no globs.
+const REPO_MARKERS: ReadonlyArray<string> = [
+  '.git',
+  'package.json',
+  'go.mod',
+  'Cargo.toml',
+];
+
+export type GrantResolution =
+  | { ok: true; root: string; source: 'project' | 'whitelist' }
+  | { ok: false; reason: string };
+
+function isUnder(child: string, parent: string): boolean {
+  return child === parent || child.startsWith(parent + sep);
+}
+
+async function exists(path: string): Promise<boolean> {
+  try {
+    await access(path, constants.F_OK);
+    return true;
+  } catch {
+    return false;
+  }
+}
+
+async function isRepoShaped(dir: string): Promise<boolean> {
+  for (const marker of REPO_MARKERS) {
+    if (await exists(`${dir}${sep}${marker}`)) return true;
+  }
+  return false;
+}
+
+// Resolves an absolute path to its grant root or refuses with a reason
+// string suitable for surfacing to the model. Pure helper — no DB writes,
+// no broker publishes. Caller persists the root on session.allowed_read_paths
+// if it wants the grant to stick.
+//
+// Arguments:
+//   sql                   — used only to read projects.path (no writes)
+//   requestedPath         — absolute path the model wants to read
+//   projectRoot           — the session's primary project root (already
+//                           realpath'd by caller). Used to short-circuit
+//                           "already in scope".
+//   whitelistRoot         — PROJECT_ROOT_WHITELIST from config (default /opt).
+//                           Walk bound for the repo-shape fallback.
+//
+// Returns { ok: true, root, source } on success; { ok: false, reason } else.
+export async function resolveGrantRoot(
+  sql: Sql,
+  requestedPath: string,
+  projectRoot: string,
+  whitelistRoot: string,
+): Promise<GrantResolution> {
+  if (typeof requestedPath !== 'string' || requestedPath.length === 0) {
+    return { ok: false, reason: 'path is required' };
+  }
+  if (!isAbsolute(requestedPath)) {
+    return { ok: false, reason: 'path must be absolute' };
+  }
+
+  // Resolve symlinks so subsequent ancestor checks compare apples-to-apples
+  // with realpath'd projectRoot. If the path doesn't exist at all, bail
+  // before bothering the user — the model is asking about a phantom.
+  let real: string;
+  try {
+    real = await realpath(requestedPath);
+  } catch {
+    return { ok: false, reason: `path does not exist: ${requestedPath}` };
+  }
+
+  // Whitelist guard. Symlinked inputs can resolve outside the whitelist
+  // even when the surface-form path looks inside it; that's why we test
+  // the *real* path here, not the requested one.
+  let realWhitelist: string;
+  try {
+    realWhitelist = await realpath(whitelistRoot);
+  } catch {
+    return { ok: false, reason: `whitelist root does not exist: ${whitelistRoot}` };
+  }
+  if (!isUnder(real, realWhitelist)) {
+    return { ok: false, reason: 'path outside permitted scope' };
+  }
+
+  // Already in scope? No prompt needed; the tool's caller should retry.
+  if (isUnder(real, projectRoot)) {
+    return { ok: false, reason: 'path already accessible without a grant' };
+  }
+
+  // Look for a registered project whose root is an ancestor of the
+  // requested path. Pick the LONGEST match (nearest ancestor wins) so
+  // sub-projects don't get over-broadened.
+  const projectRows = await sql<{ path: string }[]>`
+    SELECT path FROM projects WHERE status = 'open'
+  `;
+  let bestProject: string | null = null;
+  for (const row of projectRows) {
+    if (!row.path) continue;
+    if (!isUnder(real, row.path)) continue;
+    if (bestProject === null || row.path.length > bestProject.length) {
+      bestProject = row.path;
+    }
+  }
+  if (bestProject !== null) {
+    return { ok: true, root: bestProject, source: 'project' };
+  }
+
+  // Repo-shape fallback. Walk from the requested path upward toward the
+  // whitelist root. At every iteration: confirm we're still inside the
+  // whitelist (so a symlinked component can't slip the bound mid-walk)
+  // and confirm we haven't hit the filesystem root. The first dir with a
+  // REPO_MARKER child is the grant root.
+  let cursor = real;
+  while (true) {
+    // Don't grant the whitelist root itself — that would be far too broad.
+    if (cursor === realWhitelist) {
+      return { ok: false, reason: 'no repo-shaped ancestor found under whitelist' };
+    }
+    if (!isUnder(cursor, realWhitelist)) {
+      return { ok: false, reason: 'path outside permitted scope' };
+    }
+    const parent = dirname(cursor);
+    if (parent === cursor) {
+      // Hit filesystem root without finding a repo marker.
+      return { ok: false, reason: 'no repo-shaped ancestor found under whitelist' };
+    }
+    if (await isRepoShaped(cursor)) {
+      return { ok: true, root: cursor, source: 'whitelist' };
+    }
+    cursor = parent;
+  }
+}
--- a/apps/server/src/services/inference/error-handler.ts
+++ b/apps/server/src/services/inference/error-handler.ts
@@ -1,7 +1,14 @@
 import type { MessageMetadata, Session } from '../../types/api.js';
+import {
+  decideHtmlArtifactWrite,
+  detectHtmlArtifact,
+  deriveHtmlTitle,
+  HTML_ARTIFACT_MAX_BYTES,
+} from '../artifacts.js';
 import * as modelContext from '../model-context.js';
 import { maybeFlagForCompaction } from './payload.js';
 import { insertParts, partsFromAssistantMessage } from './parts.js';
+import type { PartInsert } from './parts.js';
 import type { InferenceContext, StreamResult, TurnArgs } from './turn.js';

 export async function handleAbortOrError(
@@ -120,17 +127,42 @@ export async function finalizeCompletion(
  // a kind='reasoning' part alongside the text.
  // TODO(v1.13.1): wrap the UPDATE above and this insertParts in a single
  // sql.begin before flipping read authority to message_parts.
-  await insertParts(
-    ctx.sql,
-    partsFromAssistantMessage({
-      content,
-      tool_calls: null,
-      reasoning: result.reasoning,
-    }).map((p) => ({
-      ...p,
-      message_id: assistantMessageId,
-    })),
-  );
+  const baseParts: PartInsert[] = partsFromAssistantMessage({
+    content,
+    tool_calls: null,
+    reasoning: result.reasoning,
+  }).map((p) => ({
+    ...p,
+    message_id: assistantMessageId,
+  }));
+  // v1.14.x-html-artifact-panes: opportunistic HTML detection. Adds a
+  // SIBLING html_artifact part — never replaces the text part. 1MB cap is
+  // graceful: oversized payloads are skipped and the assistant message
+  // lands as plain content (warn logged).
+  const htmlContent = detectHtmlArtifact(content);
+  if (htmlContent !== null) {
+    const decision = decideHtmlArtifactWrite(htmlContent);
+    if (!decision.write) {
+      ctx.log.warn(
+        { assistantMessageId, byteLen: decision.byteLen, cap: HTML_ARTIFACT_MAX_BYTES },
+        'html_artifact exceeded 1MB cap; skipping artifact part',
+      );
+    } else {
+      const title = deriveHtmlTitle(htmlContent);
+      const nextSeq = baseParts.reduce((m, p) => Math.max(m, p.sequence), -1) + 1;
+      baseParts.push({
+        message_id: assistantMessageId,
+        sequence: nextSeq,
+        kind: 'html_artifact',
+        payload: {
+          html_content: htmlContent,
+          char_count: htmlContent.length,
+          title,
+        },
+      });
+    }
+  }
+  await insertParts(ctx.sql, baseParts);
  // v1.11: flag for compaction on the terminal turn too. Catches the common
  // case of a turn that hit the limit without invoking tools.
  await maybeFlagForCompaction(ctx, chatId, updated);
--- a/apps/server/src/services/inference/index.ts
+++ b/apps/server/src/services/inference/index.ts
@@ -6,6 +6,7 @@

 export {
  createInferenceRunner,
+  MAX_STEPS,
  runAssistantTurn,
  runInference,
 } from './turn.js';
@@ -16,5 +17,8 @@ export type {
  StreamResult,
  TurnArgs,
 } from './turn.js';
+export type { ToolPhaseResult } from './tool-phase.js';
 export { detectDoomLoop, DOOM_LOOP_THRESHOLD } from './sentinels.js';
 export { buildMessagesPayload } from './payload.js';
+export { generateToolUseSummary } from './tool-summaries.js';
+export type { ToolInfo } from './tool-summaries.js';
--- a/apps/server/src/services/inference/parts.ts
+++ b/apps/server/src/services/inference/parts.ts
@@ -11,13 +11,16 @@ import type { ToolCall, ToolResult } from '../../types/api.js';
 // (schema.sql adds 'synthesis' to message_parts_kind_chk on startup). The
 // dispatch's claim that no schema migration was needed assumed kind was a
 // bare text column — it isn't; the constraint enumerates allowed values.
+// v1.14.x-html-artifact-panes: 'html_artifact' added. Schema CHECK constraint
+// in schema.sql updated in lockstep.
 export type PartKind =
  | 'text'
  | 'tool_call'
  | 'tool_result'
  | 'reasoning'
  | 'step_start'
-  | 'synthesis';
+  | 'synthesis'
+  | 'html_artifact';

 export interface PartInsert {
  message_id: string;
--- a/apps/server/src/services/inference/sentinel-summaries.ts
+++ b/apps/server/src/services/inference/sentinel-summaries.ts
@@ -476,6 +476,202 @@ export async function runDoomLoopSummary(
  );
 }

+// v1.14.0: step-cap wrap-up. Mirrors runCapHitSummary structurally — same
+// in-flight-slot reuse, same tools-disabled streaming-summary call, same
+// post-finalize sentinel insert + chat_status drop. Difference: the note
+// text names the step limit rather than the tool budget. Sentinel reuses
+// metadata.kind = 'cap_hit' so the frontend CapHitSentinel component
+// renders it without changes.
+const STEP_CAP_NOTE = (steps: number, cap: number) =>
+  `You've reached the step limit (${steps}/${cap} steps). Produce the best answer you can with what you have. Do not call more tools.`;
+
+export async function runStepCapSummary(
+  ctx: InferenceContext,
+  args: TurnArgs,
+  session: Session,
+  project: Project,
+  history: Message[],
+  agent: Agent | null,
+  steps: number,
+  cap: number,
+): Promise<void> {
+  const { sessionId, chatId, assistantMessageId, signal } = args;
+
+  const messages = await buildMessagesPayload(session, project, history, agent, ctx.log);
+  messages.push({ role: 'system', content: STEP_CAP_NOTE(steps, cap) });
+
+  const startedRow = await ctx.sql<{ started_at: string }[]>`
+    UPDATE messages
+    SET started_at = clock_timestamp()
+    WHERE id = ${assistantMessageId}
+    RETURNING started_at
+  `;
+  const startedAt = startedRow[0]?.started_at ?? null;
+
+  ctx.publish(sessionId, {
+    type: 'message_started',
+    message_id: assistantMessageId,
+    chat_id: chatId,
+    role: 'assistant',
+  });
+
+  let accumulated = '';
+  let pendingFlushTimer: NodeJS.Timeout | null = null;
+  let flushPromise: Promise<unknown> = Promise.resolve();
+  const flushNow = () => {
+    if (pendingFlushTimer) {
+      clearTimeout(pendingFlushTimer);
+      pendingFlushTimer = null;
+    }
+    const snapshot = accumulated;
+    flushPromise = flushPromise.then(() =>
+      ctx.sql`UPDATE messages SET content = ${snapshot} WHERE id = ${assistantMessageId}`
+    );
+  };
+  const scheduleFlush = () => {
+    if (pendingFlushTimer) return;
+    pendingFlushTimer = setTimeout(() => {
+      pendingFlushTimer = null;
+      flushNow();
+    }, DB_FLUSH_INTERVAL_MS);
+  };
+
+  let summaryOk = false;
+  let summarySoftCancelled = false;
+  let summaryError: string | null = null;
+  let result: StreamResult | null = null;
+  try {
+    result = await streamCompletion(
+      ctx,
+      session.model,
+      messages,
+      { tools: null, temperature: agent?.temperature },
+      (delta) => {
+        accumulated += delta;
+        ctx.publish(sessionId, {
+          type: 'delta',
+          message_id: assistantMessageId,
+          chat_id: chatId,
+          content: delta,
+        });
+        scheduleFlush();
+      },
+      undefined,
+      signal,
+    );
+    summaryOk = true;
+  } catch (err) {
+    if (err instanceof Error && err.name === 'AbortError') {
+      summarySoftCancelled = true;
+    } else {
+      summaryError = err instanceof Error ? err.message : String(err);
+    }
+  } finally {
+    if (pendingFlushTimer) {
+      clearTimeout(pendingFlushTimer);
+      pendingFlushTimer = null;
+    }
+    await flushPromise;
+  }
+
+  if (summaryOk && result) {
+    const mctx = await modelContext.getModelContext(session.model);
+    const nCtx = mctx?.n_ctx ?? null;
+    const [updated] = await ctx.sql<
+      { tokens_used: number | null; ctx_used: number | null; ctx_max: number | null; finished_at: string | null }[]
+    >`
+      UPDATE messages
+      SET content = ${result.content},
+          status = 'complete',
+          tokens_used = ${result.completionTokens},
+          ctx_used = ${result.promptTokens},
+          ctx_max = ${nCtx},
+          finished_at = clock_timestamp()
+      WHERE id = ${assistantMessageId}
+      RETURNING tokens_used, ctx_used, ctx_max, finished_at
+    `;
+    ctx.publish(sessionId, {
+      type: 'message_complete',
+      message_id: assistantMessageId,
+      chat_id: chatId,
+      tokens_used: updated?.tokens_used ?? null,
+      ctx_used: updated?.ctx_used ?? null,
+      ctx_max: updated?.ctx_max ?? null,
+      started_at: startedAt,
+      finished_at: updated?.finished_at ?? null,
+      model: session.model,
+    });
+  } else if (summarySoftCancelled) {
+    await ctx.sql`
+      UPDATE messages
+      SET content = ${accumulated},
+          status = 'cancelled',
+          finished_at = clock_timestamp()
+      WHERE id = ${assistantMessageId}
+    `;
+    ctx.publish(sessionId, {
+      type: 'message_complete',
+      message_id: assistantMessageId,
+      chat_id: chatId,
+    });
+  } else {
+    const errMeta: MessageMetadata = {
+      kind: 'error',
+      error_reason: 'summary_after_cap_failed',
+      error_text: summaryError ?? 'step-cap summary failed',
+    };
+    await ctx.sql`
+      UPDATE messages
+      SET content = ${accumulated},
+          status = 'failed',
+          finished_at = clock_timestamp(),
+          metadata = ${ctx.sql.json(errMeta as never)}
+      WHERE id = ${assistantMessageId}
+    `;
+    ctx.publish(sessionId, {
+      type: 'error',
+      message_id: assistantMessageId,
+      chat_id: chatId,
+      error: summaryError ?? 'step-cap summary failed',
+      reason: 'summary_after_cap_failed',
+    });
+  }
+
+  const [sessRow] = await ctx.sql<{ project_id: string; name: string; updated_at: string }[]>`
+    UPDATE sessions SET updated_at = clock_timestamp()
+    WHERE id = ${sessionId}
+    RETURNING project_id, name, updated_at
+  `;
+  ctx.publishUser({
+    type: 'session_updated',
+    session_id: sessionId,
+    project_id: sessRow!.project_id,
+    name: sessRow!.name,
+    updated_at: sessRow!.updated_at,
+  });
+
+  // Reuse cap_hit sentinel so the frontend CapHitSentinel component renders
+  // it without changes. The content text distinguishes step cap from budget.
+  await insertCapHitSentinel(ctx, sessionId, chatId, agent, cap);
+
+  if (summaryOk || summarySoftCancelled) {
+    ctx.publishUser({ type: 'chat_status', chat_id: chatId, status: 'idle', at: new Date().toISOString() });
+  } else {
+    ctx.publishUser({
+      type: 'chat_status',
+      chat_id: chatId,
+      status: 'error',
+      at: new Date().toISOString(),
+      reason: 'summary_after_cap_failed',
+    });
+  }
+
+  ctx.log.info(
+    { sessionId, chatId, assistantMessageId, steps, cap, summaryOk, summaryCancelled: summarySoftCancelled },
+    'inference step-cap summary finished',
+  );
+}
+
 async function insertDoomLoopSentinel(
  ctx: InferenceContext,
  sessionId: string,
--- a/apps/server/src/services/inference/stream-phase.ts
+++ b/apps/server/src/services/inference/stream-phase.ts
@@ -5,6 +5,7 @@ import type {
 } from '../../types/api.js';
 import * as modelContext from '../model-context.js';
 import { toolJsonSchemas, type ToolJsonSchema } from '../tools.js';
+import { matchToolGlob } from '../agents.js';
 import type { OpenAiMessage } from './payload.js';
 // v1.13.16: extractToolCallBlocks replaces the inline opener-search loop and
 // recognizes both Qwen <tool_call> and Anthropic <invoke> markup in one pass.
@@ -376,14 +377,14 @@ export async function executeStreamPhase(
  };

  // Tool whitelist: if an agent is set, filter the global tool list to only the
-  // tool names it allows. Unknown names in agent.tools are dropped silently
-  // (handled here by intersection). When no agent: send all tools.
+  // tool names it allows. v1.15.0-mcp-multi: uses matchToolGlob for glob
+  // pattern support (e.g. `context7_*`, `!web_*`). When no agent: send all tools.
  // v1.11.8: a second filter strips web_search + web_fetch unless the chat
  // has them explicitly enabled. Counts as an opt-in security boundary: the
  // model can't summon a tool that wasn't offered to it.
  const WEB_TOOL_NAMES: ReadonlySet<string> = new Set(['web_search', 'web_fetch']);
  const effectiveTools: ToolJsonSchema[] = (agent
-    ? toolJsonSchemas().filter((t) => agent.tools.includes(t.function.name))
+    ? toolJsonSchemas().filter((t) => matchToolGlob(t.function.name, agent.tools))
    : toolJsonSchemas()
  ).filter((t) => webToolsEnabled || !WEB_TOOL_NAMES.has(t.function.name));
  const effectiveTemperature = agent?.temperature;
--- a/apps/server/src/services/inference/tool-phase.ts
+++ b/apps/server/src/services/inference/tool-phase.ts
@@ -10,16 +10,15 @@ import { insertParts, partsFromAssistantMessage, partsFromToolMessage } from './
 // dispatch layer we no longer know which format produced the call, and the
 // extra signal is harmless for Qwen-derived calls.
 import { formatUnknownToolError } from './tool-suggestions.js';
+// v1.13.17-cross-repo-reads: pre-prompt validation for request_read_access.
+// Resolves the grant root before pausing the loop so the user is never
+// prompted about paths we couldn't grant anyway (e.g. /etc/passwd).
+import { resolveGrantRoot } from '../grant_resolver.js';
 import type {
  InferenceContext,
  StreamResult,
  TurnArgs,
 } from './turn.js';
-// v1.12.4: ESM value-import cycle. executeToolPhase recurses into
-// runAssistantTurn which lives in inference.ts. The cycle is safe because
-// the reference is read at call time (inside an async function body), not
-// at module top-level. Node + tsc resolve this cleanly.
-import { runAssistantTurn } from './turn.js';
 // v1.13.13: synthesis pipeline — replaces the immediate recursive turn when
 // any of this batch's tool calls is in SYNTHESIS_TOOLS. Falls through to
 // recursion on synthesis failure (timeout / model error). See module header
@@ -28,7 +27,8 @@ import { SYNTHESIS_TOOLS, runSynthesisPass } from '../synthesisPipeline.js';

 async function executeToolCall(
  projectRoot: string,
-  toolCall: ToolCall
+  toolCall: ToolCall,
+  extraRoots: readonly string[],
 ): Promise<{ output: unknown; truncated: boolean; error?: string }> {
  const tool = TOOLS_BY_NAME[toolCall.name];
  if (!tool) {
@@ -63,7 +63,7 @@ async function executeToolCall(
    };
  }
  try {
-    const output = await tool.execute(parsed.data, projectRoot);
+    const output = await tool.execute(parsed.data, projectRoot, extraRoots);
    const truncated =
      typeof output === 'object' && output !== null && 'truncated' in output
        ? Boolean((output as { truncated: unknown }).truncated)
@@ -81,6 +81,16 @@ async function executeToolCall(
  }
 }

+// v1.14.0: return struct from executeToolPhase so the caller (the outer
+// while loop in turn.ts) can decide whether to continue, break, or handle
+// synthesis. Replaces the recursive call into runAssistantTurn.
+export interface ToolPhaseResult {
+  action: 'continue' | 'paused' | 'synthesis_done';
+  toolCallCount: number;
+  toolCalls: ToolCall[];
+  nextAssistantId: string | null;
+}
+
 export async function executeToolPhase(
  ctx: InferenceContext,
  args: TurnArgs,
@@ -88,8 +98,8 @@ export async function executeToolPhase(
  startedAt: string | null,
  session: Session,
  projectRoot: string
-): Promise<void> {
-  const { sessionId, chatId, assistantMessageId, toolsUsed, signal } = args;
+): Promise<ToolPhaseResult> {
+  const { sessionId, chatId, assistantMessageId } = args;
  const { content, toolCalls, promptTokens, completionTokens } = result;

  // v1.11.3: ctx_max comes from llama-swap /upstream/<model>/props, not the
@@ -105,7 +115,6 @@ export async function executeToolPhase(
    UPDATE messages
    SET content = ${content},
        status = 'complete',
-        tool_calls = ${ctx.sql.json(toolCalls as never)},
        tokens_used = ${completionTokens},
        ctx_used = ${promptTokens},
        ctx_max = ${nCtx},
@@ -113,15 +122,11 @@ export async function executeToolPhase(
    WHERE id = ${assistantMessageId}
    RETURNING tokens_used, ctx_used, ctx_max, finished_at
  `;
-  // v1.13.0: dual-write to message_parts. v1.13.1-B made parts authoritative
-  // for reads via the messages_with_parts view; the JSON column write above
-  // remains for v1.13.1 fallback compatibility (dropped in v1.13.2).
+  // v1.13.20: message_parts is the sole source of truth for tool_calls.
+  // Legacy messages.tool_calls column was dropped; reads route through the
+  // messages_with_parts view.
  // v1.13.1-C: include result.reasoning so models with separate reasoning
  // channels (qwen3.6) get a kind='reasoning' part at sequence 0.
-  // TODO(v1.13.1): wrap the UPDATE above and this insertParts in a single
-  // sql.begin before flipping read authority to message_parts. Without the
-  // transaction, a crash between the two leaves an orphan message that
-  // becomes invisible in the parts-authoritative read path.
  await insertParts(
    ctx.sql,
    partsFromAssistantMessage({
@@ -187,16 +192,9 @@ export async function executeToolPhase(
      if (tc.name === 'ask_user_input') {
        pausingForUserInput = true;
        const sentinel = { tool_call_id: tc.id, output: null, truncated: false };
-        await ctx.sql`
-          UPDATE messages
-          SET tool_results = ${ctx.sql.json(sentinel as never)}
-          WHERE id = ${toolMessageId}
-        `;
-        // v1.13.0: mirror the pending sentinel into message_parts. The
-        // answer-endpoint UPDATE later (messages.ts:576) will delete and
-        // re-insert this part when the user submits their answer.
-        // TODO(v1.13.1): wrap the INSERT + UPDATE + insertParts triple in
-        // a per-iteration sql.begin before flipping read authority.
+        // v1.13.20: parts-only. The answer-endpoint UPDATE later
+        // (messages.ts) will delete and re-insert this part when the user
+        // submits their answer.
        await insertParts(
          ctx.sql,
          partsFromToolMessage({ tool_results: sentinel }).map((p) => ({
@@ -206,7 +204,63 @@ export async function executeToolPhase(
        );
        return;
      }
-      const tres = await executeToolCall(projectRoot, tc);
+      // v1.13.17-cross-repo-reads: request_read_access pauses identically to
+      // ask_user_input EXCEPT for an up-front validation pass — if the path
+      // can't be granted under the whitelist / repo-shape rules, surface an
+      // immediate denial without prompting the user. Per design D1, we never
+      // ask the user about /etc/passwd or paths outside PROJECT_ROOT_WHITELIST.
+      if (tc.name === 'request_read_access') {
+        const tcArgs = tc.args as { path?: unknown; reason?: unknown };
+        const requested =
+          typeof tcArgs.path === 'string' ? tcArgs.path : '';
+        const resolution = await resolveGrantRoot(
+          ctx.sql,
+          requested,
+          projectRoot,
+          ctx.config.PROJECT_ROOT_WHITELIST,
+        );
+        if (!resolution.ok) {
+          // Auto-deny without pausing. The model sees the reason on its
+          // next turn and decides what to do.
+          const stored = {
+            tool_call_id: tc.id,
+            output: `denied: ${resolution.reason}`,
+            truncated: false,
+          };
+          // v1.13.20: parts-only write.
+          await insertParts(
+            ctx.sql,
+            partsFromToolMessage({ tool_results: stored }).map((p) => ({
+              ...p,
+              message_id: toolMessageId,
+            })),
+          );
+          ctx.publish(sessionId, {
+            type: 'tool_result',
+            tool_message_id: toolMessageId,
+            chat_id: chatId,
+            tool_call_id: tc.id,
+            output: stored.output,
+            truncated: false,
+          });
+          return;
+        }
+        // Path is plausibly grantable — install the pending sentinel and
+        // pause. The grant endpoint re-derives the root at decision time
+        // (state may have changed in the meantime) so we don't stash it here.
+        pausingForUserInput = true;
+        const sentinel = { tool_call_id: tc.id, output: null, truncated: false };
+        // v1.13.20: parts-only write.
+        await insertParts(
+          ctx.sql,
+          partsFromToolMessage({ tool_results: sentinel }).map((p) => ({
+            ...p,
+            message_id: toolMessageId,
+          })),
+        );
+        return;
+      }
+      const tres = await executeToolCall(projectRoot, tc, session.allowed_read_paths);
      if (SYNTHESIS_TOOLS.has(tc.name)) {
        synthEntries.push({ tc, output: tres.output, ...(tres.error ? { error: tres.error } : {}) });
      }
@@ -216,14 +270,7 @@ export async function executeToolPhase(
        truncated: tres.truncated,
        ...(tres.error ? { error: tres.error } : {}),
      };
-      await ctx.sql`
-        UPDATE messages
-        SET tool_results = ${ctx.sql.json(stored as never)}
-        WHERE id = ${toolMessageId}
-      `;
-      // v1.13.0: dual-write the tool_result part.
-      // TODO(v1.13.1): wrap the INSERT + UPDATE + insertParts triple in a
-      // per-iteration sql.begin before flipping read authority.
+      // v1.13.20: parts-only write. Reads route through messages_with_parts.
      await insertParts(
        ctx.sql,
        partsFromToolMessage({ tool_results: stored }).map((p) => ({
@@ -254,7 +301,12 @@ export async function executeToolPhase(
      { sessionId, chatId, assistantMessageId },
      'inference paused awaiting user input',
    );
-    return;
+    return {
+      action: 'paused' as const,
+      toolCallCount: toolCalls.length,
+      toolCalls,
+      nextAssistantId: null,
+    };
  }

  // v1.13.13: synthesis-pipeline branch. When any of this batch's tool calls
@@ -286,30 +338,30 @@ export async function executeToolPhase(
      ...(typeof out?.truncated === 'boolean' ? { truncated: out.truncated } : {}),
      ...(typeof out?.outputPath === 'string' ? { outputPath: out.outputPath } : {}),
    });
-    if (ran) return;
+    if (ran) {
+      return {
+        action: 'synthesis_done' as const,
+        toolCallCount: toolCalls.length,
+        toolCalls,
+        nextAssistantId: null,
+      };
+    }
    // ran === false → synthesis failed (timeout / model error) → fall through
-    // to the standard recursive turn below. The synth message (if created)
+    // to the standard continue path below. The synth message (if created)
    // was already marked status='failed' inside runSynthesisPass.
  }

+  // v1.14.0: create the next assistant row and return a continue result.
+  // The caller (outer while loop in turn.ts) handles the iteration.
  const [nextAssistant] = await ctx.sql<{ id: string }[]>`
    INSERT INTO messages (session_id, chat_id, role, content, status, created_at)
    VALUES (${sessionId}, ${chatId}, 'assistant', '', 'streaming', clock_timestamp())
    RETURNING id
  `;
-  await runAssistantTurn(ctx, {
-    sessionId,
-    chatId,
-    assistantMessageId: nextAssistant!.id,
-    // v1.8.2: charge this turn's actual tool invocations against the budget.
-    // One assistant message can emit multiple tool_calls, so we add the run
-    // count, not 1. The next turn's budget check sees the cumulative total.
-    toolsUsed: toolsUsed + result.toolCalls.length,
-    // v1.11.6: append the just-executed tool calls to the per-turn history
-    // so the next runAssistantTurn's doom-loop check can see them. We don't
-    // cap the array length here — per-turn budgets keep it bounded
-    // (typically <30 entries), and slicing happens inside detectDoomLoop.
-    recentToolCalls: [...args.recentToolCalls, ...result.toolCalls],
-    signal,
-  });
+  return {
+    action: 'continue' as const,
+    toolCallCount: toolCalls.length,
+    toolCalls,
+    nextAssistantId: nextAssistant!.id,
+  };
 }
--- a/apps/server/src/services/inference/tool-summaries.ts
+++ b/apps/server/src/services/inference/tool-summaries.ts
@@ -0,0 +1,81 @@
+/**
+ * v2.0.5: Tool-use summary generation.
+ *
+ * After a batch of tool calls completes, fire a cheap LLM call to generate
+ * a "git-commit-subject-style" one-liner label describing what the tools
+ * accomplished. Ported from the Qwen Code source recon.
+ */
+import type { FastifyBaseLogger } from 'fastify';
+
+const TOOL_SUMMARY_SYSTEM_PROMPT = `Write a short summary label describing what these tool calls accomplished. Think git-commit-subject, not sentence. Past tense, most distinctive noun. Max 30 characters. Output ONLY the label.
+
+Examples:
+- Searched in auth/
+- Fixed NPE in UserService
+- Created signup endpoint
+- Read config.json
+- Ran failing tests`;
+
+const INPUT_TRUNCATE = 300;
+const MAX_SUMMARY_LENGTH = 100;
+
+export interface ToolInfo {
+  name: string;
+  input: string;
+  output: string;
+}
+
+export async function generateToolUseSummary(opts: {
+  tools: ToolInfo[];
+  llamaSwapUrl: string;
+  model: string;
+  log: FastifyBaseLogger;
+  signal?: AbortSignal;
+}): Promise<string | null> {
+  const { tools, llamaSwapUrl, model, log, signal } = opts;
+  if (tools.length === 0) return null;
+  if (signal?.aborted) return null;
+
+  const toolText = tools
+    .map(t => `Tool: ${t.name}\nInput: ${t.input.slice(0, INPUT_TRUNCATE)}\nOutput: ${t.output.slice(0, INPUT_TRUNCATE)}`)
+    .join('\n\n');
+
+  try {
+    const res = await fetch(`${llamaSwapUrl}/v1/chat/completions`, {
+      method: 'POST',
+      headers: { 'Content-Type': 'application/json' },
+      body: JSON.stringify({
+        model,
+        messages: [
+          { role: 'system', content: TOOL_SUMMARY_SYSTEM_PROMPT },
+          { role: 'user', content: toolText },
+        ],
+        max_tokens: 30,
+        temperature: 0.2,
+        stream: false,
+        chat_template_kwargs: { enable_thinking: false },
+      }),
+      signal,
+    });
+    if (!res.ok) {
+      log.debug({ status: res.status }, 'tool-summary: LLM request failed');
+      return null;
+    }
+    const data = await res.json() as { choices?: Array<{ message?: { content?: string } }> };
+    const raw = data.choices?.[0]?.message?.content?.trim() ?? '';
+    if (!raw) return null;
+    // Clean: strip quotes, "Label:" prefix, cap length
+    let cleaned = raw.split('\n')[0]?.trim() ?? '';
+    cleaned = cleaned
+      .replace(/^[-*•]\s+/, '')
+      .replace(/^["'`‘’“”]|["'`‘’“”]$/g, '')
+      .replace(/^(label|summary)\s*:\s*/i, '')
+      .trim();
+    return cleaned.length > MAX_SUMMARY_LENGTH
+      ? cleaned.slice(0, MAX_SUMMARY_LENGTH).trim()
+      : cleaned || null;
+  } catch (err) {
+    log.debug({ err: err instanceof Error ? err.message : String(err) }, 'tool-summary: error');
+    return null;
+  }
+}
--- a/apps/server/src/services/inference/turn.ts
+++ b/apps/server/src/services/inference/turn.ts
@@ -16,11 +16,9 @@ import { resolveProjectRoot } from '../path_guard.js';
 import { maybeAutoNameChat } from '../auto_name.js';
 import { getAgentById } from '../agents.js';
 import * as compaction from '../compaction.js';
-import * as modelContext from '../model-context.js';
 import type { Broker } from '../broker.js';
 import { resolveToolBudget } from './budget.js';
 import {
-  DOOM_LOOP_THRESHOLD,
  detectDoomLoop,
 } from './sentinels.js';
 import {
@@ -33,15 +31,23 @@ import {
 } from './error-handler.js';
 import {
  executeStreamPhase,
-  streamCompletion,
 } from './stream-phase.js';
-import { executeToolPhase } from './tool-phase.js';
-import { DB_FLUSH_INTERVAL_MS, type StreamPhaseState } from './types.js';
+import { executeToolPhase, type ToolPhaseResult } from './tool-phase.js';
+import type { StreamPhaseState } from './types.js';
 import {
  runCapHitSummary,
  runDoomLoopSummary,
+  runStepCapSummary,
 } from './sentinel-summaries.js';

+// v1.14.0: hard ceiling on the number of stream-and-tool iterations per
+// user-message turn. Per-agent cap via agent.steps is the primary knob;
+// MAX_STEPS is the safety ceiling. 200 is 4x the effective budget ceiling
+// (50 tool calls) — in practice budget fires first unless the model makes
+// many 0-tool-call iterations (which exit the loop via the non-tool finish
+// path anyway).
+export const MAX_STEPS = 200;
+
 // v1.12.4: re-exported so external callers (tests, future consumers) keep
 // importing from services/inference.js as the public surface.
 export { detectDoomLoop, DOOM_LOOP_THRESHOLD } from './sentinels.js';
@@ -145,75 +151,185 @@ export async function runAssistantTurn(
  ctx: InferenceContext,
  args: TurnArgs,
 ): Promise<void> {
-  const { sessionId, chatId } = args;
+  const { sessionId, chatId, signal } = args;

-  // v1.11: if the prior turn flagged this chat for compaction, run it first
-  // so loadContext below reads the post-compaction history. We swallow
-  // compaction failures (clearing the flag so we don't loop) and proceed
-  // with the un-compacted history — a slow turn that hits the model's
-  // hard limit is recoverable; a dead session is not.
-  const chatFlag = await ctx.sql<{ needs_compaction: boolean }[]>`
-    SELECT needs_compaction FROM chats WHERE id = ${chatId}
-  `;
-  if (chatFlag[0]?.needs_compaction) {
-    try {
-      await compaction.process({
-        sql: ctx.sql,
-        config: ctx.config,
-        log: ctx.log,
-        broker: ctx.broker,
-        chatId,
-      });
-    } catch (err) {
-      ctx.log.warn({ err, chatId }, 'auto-compaction failed; clearing flag and proceeding');
-      await ctx.sql`UPDATE chats SET needs_compaction = false WHERE id = ${chatId}`;
-    }
-  }
-
-  const loaded = await loadContext(ctx.sql, sessionId, chatId);
-  if (!loaded) {
+  // v1.14.0: resolve agent once at the top. The agent stays fixed for the
+  // duration of this user-message turn — PATCH agent_id mid-conversation
+  // takes effect on the next runInference, not mid-loop.
+  const initialLoaded = await loadContext(ctx.sql, sessionId, chatId);
+  if (!initialLoaded) {
    ctx.log.warn({ sessionId }, 'inference: session or project missing');
    return;
  }
-  const { session, project, history } = loaded;
-  const projectRoot = await resolveProjectRoot(project.path);
-  // Agent resolution is per-turn so PATCH agent_id mid-conversation takes
-  // effect on the next message. Unknown agent_id returns null silently —
-  // session falls back to base prompt + all tools + default temperature.
+  const { session, project } = initialLoaded;
  const agent = session.agent_id
    ? await getAgentById(project.path, session.agent_id)
    : null;
-
-  // v1.8.2: cap-hit replaces the older "tool loop depth exceeded" failure.
-  // When we've already burned the budget *before* this turn even runs, we
-  // skip straight to the summary flow — the in-flight assistant message slot
-  // gets reused for the wrap-up reply instead of being marked failed.
  const budget = resolveToolBudget(agent);
-  if (args.toolsUsed >= budget) {
-    await runCapHitSummary(ctx, args, session, project, history, agent, budget);
+
+  // v1.14.0: effectiveCap = min(agent.steps ?? Infinity, MAX_STEPS).
+  // steps: 0 means "no tool calls allowed" — the first stream phase runs
+  // but if it emits tool calls they are not executed (finalize as text-only).
+  const effectiveCap = Math.min(agent?.steps ?? Infinity, MAX_STEPS);
+
+  // steps: 0 special case — model responds text-only. The while loop would
+  // never enter (effectiveCap === 0), so we handle it explicitly before the
+  // loop. The model always gets at least one chance to respond with text.
+  if (effectiveCap === 0) {
+    const loaded = await loadContext(ctx.sql, sessionId, chatId);
+    if (loaded) {
+      await runTextOnlyTurn(ctx, args, loaded.session, loaded.project, loaded.history, agent);
+    }
    return;
  }

-  // v1.11.6: doom-loop guard. Detected BEFORE the budget cap (the model can
-  // burn through 3 identical calls long before the 15-call budget fires).
-  // Same in-flight-slot-reuse pattern as runCapHitSummary — wrap-up reply
-  // lands in args.assistantMessageId, then a doom_loop sentinel is inserted
-  // to make the abort visible in the chat history.
-  const loop = detectDoomLoop(args.recentToolCalls);
-  if (loop) {
-    await runDoomLoopSummary(ctx, args, session, project, history, agent, loop);
-    return;
+  let stepNumber = 0;
+  let toolsUsed = args.toolsUsed;
+  let recentToolCalls = args.recentToolCalls;
+  let assistantMessageId = args.assistantMessageId;
+
+  while (stepNumber < effectiveCap) {
+    // ---- doom-loop check (moved from top-of-function) ----
+    const loop = detectDoomLoop(recentToolCalls);
+    if (loop) {
+      // Need fresh history for the summary.
+      const loaded = await loadContext(ctx.sql, sessionId, chatId);
+      if (loaded) {
+        const iterArgs: TurnArgs = { sessionId, chatId, assistantMessageId, toolsUsed, recentToolCalls, signal };
+        await runDoomLoopSummary(ctx, iterArgs, loaded.session, loaded.project, loaded.history, agent, loop);
+      }
+      break;
+    }
+
+    // ---- budget check (moved from top-of-function) ----
+    if (toolsUsed >= budget) {
+      const loaded = await loadContext(ctx.sql, sessionId, chatId);
+      if (loaded) {
+        const iterArgs: TurnArgs = { sessionId, chatId, assistantMessageId, toolsUsed, recentToolCalls, signal };
+        await runCapHitSummary(ctx, iterArgs, loaded.session, loaded.project, loaded.history, agent, budget);
+      }
+      break;
+    }
+
+    // ---- compaction check ----
+    // v1.11: if the prior turn flagged this chat for compaction, run it
+    // before loadContext so we read post-compaction history. Swallow
+    // failures and proceed with un-compacted history.
+    const chatFlag = await ctx.sql<{ needs_compaction: boolean }[]>`
+      SELECT needs_compaction FROM chats WHERE id = ${chatId}
+    `;
+    if (chatFlag[0]?.needs_compaction) {
+      try {
+        await compaction.process({
+          sql: ctx.sql,
+          config: ctx.config,
+          log: ctx.log,
+          broker: ctx.broker,
+          chatId,
+        });
+      } catch (err) {
+        ctx.log.warn({ err, chatId }, 'auto-compaction failed; clearing flag and proceeding');
+        await ctx.sql`UPDATE chats SET needs_compaction = false WHERE id = ${chatId}`;
+      }
+    }
+
+    // ---- load context (must re-load each iteration — new messages since last step) ----
+    const loaded = await loadContext(ctx.sql, sessionId, chatId);
+    if (!loaded) {
+      ctx.log.warn({ sessionId }, 'inference: session or project missing mid-loop');
+      break;
+    }
+    const { session: iterSession, project: iterProject, history } = loaded;
+    const projectRoot = await resolveProjectRoot(iterProject.path);
+
+    // v1.14.0: log step boundary for instrumentation. step_start parts are in
+    // the schema CHECK but not emitted here — writing to the assistant message
+    // before the stream phase creates a sequence-0 collision with
+    // partsFromAssistantMessage. A WS frame or structured log is sufficient
+    // since the frontend doesn't render step boundaries in v1.14.
+    ctx.log.info({ sessionId, chatId, step: stepNumber, assistantMessageId }, 'step_start');
+
+    // ---- build messages + stream phase ----
+    const messages = await buildMessagesPayload(iterSession, iterProject, history, agent, ctx.log);
+    const webToolsEnabled =
+      iterSession.web_search_enabled ?? iterProject.default_web_search_enabled ?? false;
+
+    const iterArgs: TurnArgs = { sessionId, chatId, assistantMessageId, toolsUsed, recentToolCalls, signal };
+    const state: StreamPhaseState = { accumulated: '', startedAt: null };
+    let result: StreamResult;
+    try {
+      result = await executeStreamPhase(ctx, iterArgs, iterSession, messages, state, agent, webToolsEnabled);
+    } catch (err) {
+      await handleAbortOrError(ctx, iterArgs, state.accumulated, err);
+      break;
+    }
+
+    // ---- non-tool finish → finalize and exit ----
+    if (result.toolCalls.length === 0) {
+      await finalizeCompletion(ctx, iterArgs, result, state.startedAt, iterSession);
+      break;
+    }
+
+    // ---- steps: 0 edge case ----
+    // effectiveCap check above guarantees we're inside the loop, but this
+    // guard handles the theoretical case where the model emits tool calls
+    // on step 0 when effectiveCap would have been 0 (impossible since the
+    // while condition prevents entry, but kept for safety). If effectiveCap
+    // is 1 and we're on step 0, tool calls ARE executed — steps counts
+    // iterations, not post-first-stream.
+
+    // ---- tool phase ----
+    let toolPhaseResult: ToolPhaseResult;
+    try {
+      toolPhaseResult = await executeToolPhase(ctx, iterArgs, result, state.startedAt, iterSession, projectRoot);
+    } catch (err) {
+      // Tool phase errors are unexpected (individual tool failures are
+      // caught inside executeToolPhase). Log and break.
+      ctx.log.error({ err, sessionId, chatId, step: stepNumber }, 'tool phase threw unexpectedly');
+      break;
+    }
+
+    // ---- update loop locals ----
+    toolsUsed += toolPhaseResult.toolCallCount;
+    recentToolCalls = [...recentToolCalls, ...toolPhaseResult.toolCalls];
+    stepNumber++;
+
+    if (toolPhaseResult.action !== 'continue') {
+      // 'paused' (user input) or 'synthesis_done' — stop the loop.
+      break;
+    }
+    // 'continue' — advance to next assistant message.
+    assistantMessageId = toolPhaseResult.nextAssistantId!;
  }

+  // ---- post-loop: step-cap sentinel ----
+  // When the loop exits because stepNumber reached effectiveCap, the last
+  // iteration's tool phase returned 'continue' with a nextAssistantId that
+  // is still in 'streaming' status (unfilled). Use it for the wrap-up.
+  if (stepNumber >= effectiveCap && effectiveCap < Infinity) {
+    const loaded = await loadContext(ctx.sql, sessionId, chatId);
+    if (loaded) {
+      const capArgs: TurnArgs = { sessionId, chatId, assistantMessageId, toolsUsed, recentToolCalls, signal };
+      await runStepCapSummary(ctx, capArgs, loaded.session, loaded.project, loaded.history, agent, stepNumber, effectiveCap);
+    }
+  }
+}
+
+// v1.14.0: special handling for steps: 0 — the model responds text-only.
+// The while loop never enters (effectiveCap === 0). We stream once with
+// no tools, finalize, and return. If the model emits tool calls despite
+// not being offered tools, they're ignored (finalize as text-only).
+async function runTextOnlyTurn(
+  ctx: InferenceContext,
+  args: TurnArgs,
+  session: Session,
+  project: Project,
+  history: Message[],
+  agent: Agent | null,
+): Promise<void> {
  const messages = await buildMessagesPayload(session, project, history, agent, ctx.log);
-
-  // v1.11.8: resolve per-chat web-tools opt-in. Tri-state on the wire:
-  //   - session.web_search_enabled = null → inherit project default
-  //   - session.web_search_enabled = true/false → explicit
-  // Both web_search and web_fetch are gated by this single flag (the UI
-  // label is "Enable web search and fetch" — same store, both tools).
-  // Default is false unless explicitly opted in, matching the v1.9
-  // plumbing intent ("inert until Batch 8 ships the actual tools").
+  // Web tools are irrelevant when steps: 0 (no tool execution), but we
+  // still need to resolve the flag for executeStreamPhase's signature.
  const webToolsEnabled =
    session.web_search_enabled ?? project.default_web_search_enabled ?? false;

@@ -227,8 +343,12 @@ export async function runAssistantTurn(
  }

  if (result.toolCalls.length > 0) {
-    await executeToolPhase(ctx, args, result, state.startedAt, session, projectRoot);
-    return;
+    ctx.log.warn(
+      { chatId: args.chatId, toolCallCount: result.toolCalls.length },
+      'steps: 0 agent emitted tool calls; ignoring and finalizing as text-only',
+    );
+    // Override: strip tool calls so finalizeCompletion treats it as text-only.
+    result = { ...result, toolCalls: [] };
  }

  await finalizeCompletion(ctx, args, result, state.startedAt, session);
--- a/apps/server/src/services/mcp-client.ts
+++ b/apps/server/src/services/mcp-client.ts
@@ -0,0 +1,288 @@
+/**
+ * v1.15.0-mcp-multi: multi-server MCP client registry.
+ *
+ * Connects to multiple MCP servers (Streamable HTTP or stdio transport),
+ * discovers tools from each, wraps them as BooCode ToolDefs with a
+ * `<serverName>_<toolName>` name prefix, and routes callTool by prefix.
+ *
+ * Graceful degradation: one failing server doesn't block others.
+ * Read-only invariant: tools with readOnlyHint === false are rejected.
+ */
+import { Client } from '@modelcontextprotocol/sdk/client';
+import { StreamableHTTPClientTransport } from '@modelcontextprotocol/sdk/client/streamableHttp.js';
+import { StdioClientTransport } from '@modelcontextprotocol/sdk/client/stdio.js';
+import { z } from 'zod';
+import type { FastifyBaseLogger } from 'fastify';
+import type { McpServerEntry, McpServerConfig } from './mcp-config.js';
+import type { ToolDef } from './tools.js';
+
+// ---- Types ----
+
+interface McpToolAnnotations {
+  readOnlyHint?: boolean;
+  destructiveHint?: boolean;
+  [key: string]: unknown;
+}
+
+interface McpToolDef {
+  name: string;
+  description?: string;
+  inputSchema: Record<string, unknown>;
+  annotations?: McpToolAnnotations;
+}
+
+interface ServerState {
+  client: Client;
+  transport: StreamableHTTPClientTransport | StdioClientTransport;
+  tools: ToolDef<Record<string, unknown>>[];
+  type: 'streamableHttp' | 'stdio';
+}
+
+// ---- Module-level state ----
+
+const servers = new Map<string, ServerState>();
+// Reverse map: prefixed tool name → server name (built during discovery)
+const toolToServer = new Map<string, string>();
+let log: FastifyBaseLogger | null = null;
+
+const MAX_RESULT_BYTES = 5 * 1024 * 1024;
+
+// ---- Public API ----
+
+/**
+ * Connect to all configured MCP servers, discover tools, and wrap them.
+ * Per-server graceful degradation: a failing server is logged and skipped.
+ */
+export async function initialize(
+  entries: McpServerEntry[],
+  logger: FastifyBaseLogger,
+): Promise<void> {
+  log = logger;
+
+  // Connect servers in parallel — each wrapped in try/catch for isolation
+  await Promise.all(
+    entries.map(async (entry) => {
+      try {
+        await connectServer(entry);
+      } catch (err) {
+        log!.warn(
+          { err, server: entry.name },
+          `mcp: failed to initialize server "${entry.name}" — its tools will be unavailable`,
+        );
+      }
+    }),
+  );
+
+  if (servers.size > 0) {
+    const totalTools = Array.from(servers.values()).reduce((n, s) => n + s.tools.length, 0);
+    log.info(
+      { servers: servers.size, tools: totalTools },
+      'mcp: multi-server initialization complete',
+    );
+  }
+}
+
+/**
+ * Call an MCP tool by its prefixed name. Routes to the correct server
+ * using the toolToServer reverse map.
+ */
+export async function callTool(
+  prefixedName: string,
+  args: Record<string, unknown>,
+): Promise<unknown> {
+  const serverName = toolToServer.get(prefixedName);
+  if (!serverName) {
+    return { error: true, output: `MCP tool "${prefixedName}" not found in any server` };
+  }
+
+  const state = servers.get(serverName);
+  if (!state) {
+    return { error: true, output: `MCP server "${serverName}" not available` };
+  }
+
+  // Strip the "<serverName>_" prefix to get the original tool name
+  const originalName = prefixedName.slice(serverName.length + 1);
+
+  try {
+    const result = await state.client.callTool({ name: originalName, arguments: args });
+
+    const content = result.content as Array<{ type: string; text?: string; [key: string]: unknown }>;
+    if (!content || content.length === 0) {
+      return '(no output)';
+    }
+
+    if (result.isError) {
+      const joined = content
+        .map((block) => (block.type === 'text' ? block.text ?? '' : JSON.stringify(block)))
+        .join('\n');
+      return { error: true, output: joined || '(MCP error with no details)' };
+    }
+
+    const parts = content.map((block) => {
+      if (block.type === 'text') return block.text ?? '';
+      return JSON.stringify(block);
+    });
+    const joined = parts.join('\n');
+    if (joined.length > MAX_RESULT_BYTES) {
+      log?.warn({ tool: originalName, server: serverName, bytes: joined.length, cap: MAX_RESULT_BYTES }, 'mcp: result truncated');
+      return joined.slice(0, MAX_RESULT_BYTES) + '\n\n[truncated — MCP result exceeded size limit]';
+    }
+    return joined;
+  } catch (err) {
+    log?.warn({ err, tool: originalName, server: serverName }, 'mcp: callTool failed');
+    return {
+      error: true,
+      output: err instanceof Error ? err.message : 'MCP server unreachable',
+    };
+  }
+}
+
+/** Return all wrapped ToolDefs from all connected servers, flattened. */
+export function getTools(): ToolDef<Record<string, unknown>>[] {
+  const all: ToolDef<Record<string, unknown>>[] = [];
+  for (const state of servers.values()) {
+    all.push(...state.tools);
+  }
+  return all;
+}
+
+/** Return status of each server (for debug/status endpoints). */
+export function getMcpServers(): Array<{
+  name: string;
+  type: 'streamableHttp' | 'stdio';
+  toolCount: number;
+  connected: boolean;
+}> {
+  return Array.from(servers.entries()).map(([name, state]) => ({
+    name,
+    type: state.type,
+    toolCount: state.tools.length,
+    connected: true,
+  }));
+}
+
+/**
+ * Graceful shutdown. For stdio servers, the SDK's transport.close() handles
+ * SIGTERM + timeout. For HTTP servers, close the transport.
+ */
+export async function shutdown(): Promise<void> {
+  const closePromises: Promise<void>[] = [];
+  for (const [name, state] of servers) {
+    closePromises.push(
+      (async () => {
+        try {
+          await state.transport.close();
+          log?.info({ server: name }, 'mcp: server transport closed');
+        } catch (err) {
+          log?.warn({ err, server: name }, 'mcp: error closing server transport');
+        }
+      })(),
+    );
+  }
+  await Promise.all(closePromises);
+  servers.clear();
+  toolToServer.clear();
+}
+
+// ---- Internal helpers ----
+
+async function connectServer(entry: McpServerEntry): Promise<void> {
+  const { name, config } = entry;
+
+  const client = new Client({ name: 'boocode', version: '1.15.0' });
+  let transport: StreamableHTTPClientTransport | StdioClientTransport;
+
+  if (config.type === 'streamableHttp') {
+    transport = createHttpTransport(config);
+  } else {
+    transport = createStdioTransport(config);
+  }
+
+  await client.connect(transport);
+
+  const result = await client.listTools();
+  const mcpTools = (result.tools ?? []) as McpToolDef[];
+
+  const tools: ToolDef<Record<string, unknown>>[] = [];
+  for (const t of mcpTools) {
+    if (t.annotations?.readOnlyHint === false) {
+      log!.info({ tool: t.name, server: name }, 'mcp: skipping non-read-only tool');
+      continue;
+    }
+    const wrapped = wrapMcpTool(name, t);
+    tools.push(wrapped);
+    toolToServer.set(wrapped.name, name);
+  }
+
+  servers.set(name, { client, transport, tools, type: config.type });
+
+  log!.info(
+    { server: name, type: config.type, count: tools.length, names: tools.map((t) => t.name) },
+    'mcp: server initialized',
+  );
+}
+
+function createHttpTransport(config: Extract<McpServerConfig, { type: 'streamableHttp' }>): StreamableHTTPClientTransport {
+  const requestInit: RequestInit = {};
+  if (config.headers && Object.keys(config.headers).length > 0) {
+    requestInit.headers = config.headers;
+  }
+  return new StreamableHTTPClientTransport(new URL(config.url), { requestInit });
+}
+
+function createStdioTransport(config: Extract<McpServerConfig, { type: 'stdio' }>): StdioClientTransport {
+  return new StdioClientTransport({
+    command: config.command,
+    args: config.args,
+    env: config.env,
+    stderr: 'pipe',
+  });
+}
+
+/** Wrap an MCP tool as a BooCode ToolDef with a server-name prefix. */
+export function wrapMcpTool(
+  serverName: string,
+  mcpTool: McpToolDef,
+): ToolDef<Record<string, unknown>> {
+  const prefixedName = `${serverName}_${mcpTool.name}`;
+  return {
+    name: prefixedName,
+    description: mcpTool.description ?? '',
+    inputSchema: z.record(z.unknown()),
+    jsonSchema: {
+      type: 'function' as const,
+      function: {
+        name: prefixedName,
+        description: mcpTool.description ?? '',
+        parameters: mcpTool.inputSchema ?? { type: 'object', properties: {} },
+      },
+    },
+    execute: async (input) => {
+      return callTool(prefixedName, input);
+    },
+  };
+}
+
+/** Exposed for unit tests — extract content from an MCP result. */
+export function extractContent(
+  content: Array<{ type: string; text?: string; [key: string]: unknown }> | undefined,
+  isError?: boolean,
+): unknown {
+  if (!content || content.length === 0) return '(no output)';
+
+  const parts = content.map((block) => {
+    if (block.type === 'text') return block.text ?? '';
+    return JSON.stringify(block);
+  });
+  const joined = parts.join('\n');
+
+  if (isError) {
+    return { error: true, output: joined || '(MCP error with no details)' };
+  }
+  return joined;
+}
+
+/** Exposed for unit tests — the read-only guard predicate. */
+export function isToolReadOnly(annotations?: McpToolAnnotations): boolean {
+  return annotations?.readOnlyHint !== false;
+}
--- a/Show More
+++ b/Show More