knowledge-rag

What is this MCP

This is a local Retrieval-Augmented Generation (RAG) system designed specifically for Claude Code. It provides a complete document search and retrieval solution that runs entirely locally without requiring external servers or API keys. The system features hybrid search combining semantic and BM25 retrieval, cross-encoder reranking, markdown-aware chunking, and supports 9 different file formats.

How to use this MCP

Install the package via pip (pip install knowledge-rag) and integrate it with Claude Code. The system includes 12 MCP tools that allow you to drop documents into a watched directory, which are then automatically parsed, chunked, and indexed. You can search through your documents using natural language queries, and the system will return relevant chunks with hybrid search and reranking. The file watcher monitors your document directory for changes and updates the index automatically.

What this MCP can be used for

This MCP is ideal for developers and researchers who need to search through technical documentation, codebases, research papers, or any collection of documents while working within Claude Code. It enables efficient knowledge retrieval from local document collections, supports 20 different format parsers for various file types, and provides high-quality search results through its hybrid search and reranking capabilities. The 100% local operation ensures data privacy and eliminates dependency on external services.