basemind-code-search

basemind-code-search — navigate code without reading it

basemind pre-indexes the repo into a tree-sitter code map across 300+ languages. Structural questions — where a symbol lives, what calls it, what shape a file has — resolve from the index in milliseconds and return paths, line numbers, and signatures, not file bodies. That is a fraction of the tokens of reading source, so it is the default, not an optimization.

basemind first, grep/read fallback. If a question is about where, what calls, what shape, or what implements, a basemind tool answers it cheaper than grep/rg or opening files. Drop to raw shell only when no tool covers the question.

The discipline

• outline a file before you open it. A 1000-line file becomes a 30-line table of contents. Read the actual source only once you have the exact span, then read that range, not the file.
• search_symbols instead of grep for a definition. It matches indexed symbol names and returns path:line, skipping the comment/string/test-name noise grep drowns you in.
• find_references / find_callers instead of grepping call sites. Indexed call edges, not text matches.
• workspace_grep instead of shelling out to ripgrep when you genuinely need regex over content — it runs over the in-RAM index and returns capped, structured hits.
• Do not re-read a file basemind already mapped. If the outline answered the question, stop.
• rescan after you edit code, not a server reconnect. Pass paths: [...] to limit it.

Tool routing

Question	MCP tool	CLI
"Where is X defined?"	search_symbols (substring, optional kind)	basemind query symbol "X"
"What's the shape of file F?"	outline (add l2: true for calls + docs)	basemind query outline F [--l2]
"What calls X?" (any name)	find_references	basemind query references "X"
"What calls this specific definition?"	find_callers (path + name + optional kind)	basemind query callers F name [--kind]
"Trace the call graph from a function?"	call_graph (BFS, bounded by max_depth / max_nodes)	basemind query call-graph "name" [--direction --max-depth]
"What implements / extends / inherits X?"	find_implementations (Rust, Python, TS/TSX, JS)	basemind query implementations "X"
"What imports module M?"	dependents	basemind query dependents "M"
"What files are indexed?"	list_files (filter by language / path_contains)	basemind query list-files [--language --path-contains]
"Regex over file contents?"	workspace_grep	basemind query grep "pattern" [--language --path-contains]
"What's indexed?"	status	basemind query status
"Refresh the index after editing?"	rescan (optional paths)	basemind scan
"Fetch the next page?"	pass next_cursor from the prior response as cursor	—

Examples

search_symbols { needle: "MapCache" }
→ src/mcp/mod.rs:79:1 MapCache (struct)
  src/mcp/mod.rs:88:1 MapCache (impl)

find_references { name: "process_file" }
→ src/scanner.rs:142:9 process_file
  src/scanner.rs:201:13 process_file

outline { path: "src/mcp/tools.rs" }
→ 21 #[tool] outline (function)
  112 #[tool] search_symbols (function)

Notes

• Matching on symbol names is substring: find_references("bar") matches Foo::bar() and bar() alike. There is no scope resolution — cross-check with outline when disambiguation matters.
• Lists are capped (limit, default 100, max 1000). Index scanners use scan_cap = limit * 8 to bound work on common names.
• Needs an index at .basemind/ — run basemind scan first (see the basemind-scan skill). "No indexed files" means the scan hasn't run in this repo yet.

For git history / blame / diffs see basemind-git-history; for document RAG and semantic search see basemind-documents; for agent coordination see basemind-comms.