brainclaw privacy

Analytics only if you say yes.

The site works without analytics. If you opt in, we use analytics to understand what helps and what confuses people. No ads. No resale. No dark patterns.

Privacy details

🗺️

Code Map

Your agent knows where things live before it reads a single file.

soloteambuilder code-mapcontext

A fresh agent — or you at 2am in an unfamiliar repo — burns the first ten minutes grepping for where getLocalizedPath actually lives and what touches it. Code Map removes that tax. brainclaw parses your code with Tree-sitter — JavaScript, TypeScript/TSX, Python, PHP and Java in one index — and records the symbols each file defines (functions, classes, types, interfaces, React components and hooks), what it imports and exports, and how files relate — then answers “what should I read before I edit this?” in one call.

It’s a discovery aid, not a build artifact: it never changes your code, never blocks bclaw_work, and the index (JSONL shards under .brainclaw/code/) is safe to delete — a refresh rebuilds it.

brainclaw code-map status                     # coverage + freshness, never parses
brainclaw code-map refresh                    # incremental: only changed files
brainclaw code-map find getLocalizedPath
#   [10.0] getLocalizedPath  function — src/i18n/config.ts
brainclaw code-map brief src/i18n/config.ts
#   → ranked files to read next (max 12) + the decisions and traps attached

Never silently wrong

Code Map is pull-based — no daemon, no background reindex. Every status / find / brief recomputes a freshness badge from git status and file hashes, so a stale index is always visible instead of quietly misleading:

BadgeMeaning
freshMatches the working tree, the extractor config, and the parser binaries
stale_changed_filesIndexed files changed on disk since they were parsed
stale_extractor / stale_grammarThe config or a Tree-sitter grammar was bumped
missing_indexNo index exists yet for this project

refresh --changed (the default) re-parses only content-changed files and heals stale_extractor / stale_grammar on that same cheap path; --all does a full rebuild with orphan compaction. A held project lock makes refresh fail fast — it never stalls your work, so the worst case is a one-line “run refresh” hint, never a wrong answer. The Tree-sitter parser ships as bundled WASM and loads only on the first parse, so status / find / brief stay instant even before the engine is touched.

On the MCP server, and per-project

The same four calls are MCP tools — bclaw_code_status, bclaw_code_find, bclaw_code_brief, bclaw_code_refresh — so a capable agent reaches them mid-task without a shell, each carrying the same freshness badge; the read tools never trigger a parse. In a monorepo the index is per project: run it inside apps/api and you get that child’s own map. brainclaw’s project resolution routes Code Map to the project you’re working in — the same scoping that powers bclaw_work and bclaw_switch — with no --cwd juggling.

And because brainclaw already holds your decisions and traps, brief doesn’t just point at code — it surfaces the memory attached to that scope. Ask about a module and you get the files to read next and the trap someone left there last month.