LeanKG maps repos for coding agents

LeanKG is aimed at a real pain point in agentic coding, but its strongest pitch is not just token savings; it is turning codebase navigation into structured graph queries.

• –Local indexing and MCP integration make it more practical for private repos than cloud-only code intelligence tools.
• –Tree-sitter-based parsing gives it a better foundation than naive text search for impact analysis, signatures, and dependency traversal.
• –The benchmark story needs nuance: the repo advertises up to 98% savings for impact analysis, while its own broader AB notes show mixed results and pending deduplication work.
• –The category is getting crowded fast, so LeanKG’s install hooks across Cursor, Claude Code, OpenCode, Gemini, and Codex may matter as much as the graph engine itself.