Codebase RAG

Provides semantic codebase search and file retrieval using CocoIndex embeddings and PostgreSQL with pgvector for similarity search, enabling natural language queries to find relevant code snippets with similarity scores and line numbers.

This MCP server provides semantic codebase search and file retrieval using CocoIndex embeddings and PostgreSQL with pgvector for similarity search. Built by allentcm, it integrates with the Voyage API for code-specific embeddings (voyage-code-3 model) and offers three core tools: semantic code search that returns relevant snippets with similarity scores and line numbers, file listing from the tracking table, and complete file content retrieval. The implementation uses CocoIndex's transform flow system for embedding generation, supports Docker deployment with environment-based configuration, and is designed for AI assistants that need to understand and navigate large codebases through natural language queries rather than traditional keyword-based search.