LeanKG
View on GitHubLeanKG: Stop Burning Tokens. Start Coding Lean.
Go MCP server that indexes a codebase into a knowledge graph (SQLite by default, optional Postgres/pgvector) and exposes 3 tools / 30 actions so AI coding agents get targeted subgraphs instead of grep-and-read context, cutting tokens ~65% and tool calls ~85%.
Use Cases
Cut token usage in AI coding agentsCodebase exploration over MCPImpact/blast-radius analysis before refactorsCall graph and dependency queriesSemantic code searchPRD-to-code traceabilityMulti-repo indexing for teamsOnboarding to unfamiliar codebasesService and incident context lookupWire code knowledge into Claude Code, Cursor, Codex, Gemini CLIIndex codebase in CI with zero-config SQLite
Built With
- Language
- Go
- Frameworks
- Model Context Protocol (Go SDK) · go-tree-sitter · pgvector · ConnectRPC · Protocol Buffers · SQLite · PostgreSQL · fsnotify · Vite · React
Tags
mcp · mcp-server · knowledge-graph · code-graph · code-search · semantic-search · context-engineering · token-reduction · ai-memory · tree-sitter · embeddings · blast-radius · call-graph · multi-repo · sqlite · golang