Code Graph RAG

Provides advanced code graph analysis through specialized agents for parsing, indexing, querying, and semantic search across codebases with Tree-sitter parsing for multiple languages, SQLite vector storage, and hybrid search combining structural graph traversal with semantic similarity for code review workflows, refactoring analysis, and natural language codebase understanding.

This multi-agent LiteRAG MCP server provides advanced code graph analysis through specialized agents for parsing, indexing, querying, and semantic search across codebases. Built with TypeScript and featuring Tree-sitter parsing for multiple languages (TypeScript, JavaScript, Python, C/C++), SQLite with sqlite-vec for vector storage, and optional embedding models, it offers 13 MCP tools including semantic search, code similarity detection, clone analysis, impact assessment, and cross-language pattern matching. The implementation uses an agent-based architecture with a coordinator for task distribution, supports incremental parsing for large codebases, includes comprehensive caching and performance optimization for commodity hardware, and provides hybrid search combining structural graph traversal with semantic vector similarity, making it valuable for code review workflows, refactoring analysis, and understanding complex codebases through natural language queries.