knowledge-rag

Local RAG server for Claude Code, Cursor and other MCP clients: hybrid BM25 + vector search with cross-encoder reranking on by default, 13 MCP tools, 33 file ty

What is knowledge-rag?

Local RAG server for Claude Code, Cursor and other MCP clients: hybrid BM25 + vector search with cross-encoder reranking on by default, 13 MCP tools, 33 file types including code, re-indexing on file changes. One Python process with embedded ChromaDB and FastEmbed ONNX: no Docker, database server or API key; offline after the first model download.

Overview of knowledge-rag

What it is

knowledge-rag is an MCP-native local RAG server for AI agents such as Claude Code, Cursor, Windsurf, VS Code, Cline, Gemini CLI and Zed. It performs hybrid search (semantic + BM25), cross-encoder reranking, and provides 13 MCP tools. It parses 35 file formats. The server runs entirely locally with no external servers or API keys required. It includes optional bearer authentication, Prometheus metrics, rate limiting, health probes, structured JSON logging and staging-based rebuilds. The project is licensed under MIT and is open source.

Key features

  • Hybrid search combining semantic embeddings and BM25 retrieval - Cross-encoder reranking for improved result relevance - 13 MCP tools for integration with AI clients - Parses 35 file formats (PDF, markdown, code, etc.)
  • 100% local processing: embedding generation, indexing, and retrieval run on-device - Zero external servers or API keys; works offline after first model download - Production-grade features: bearer auth, Prometheus metrics, rate limiting, health probes, structured JSON logging

Use cases

  • Local document Q&A and retrieval for AI coding assistants (Claude Code, Cursor, etc.)
  • Knowledge base search over proprietary or sensitive documents without sending data to the cloud - Multi-user shared RAG server via HTTP/SSE transport with authentication and rate limiting - Air-gapped or offline deployments using Docker with pre-downloaded models

Getting started

bash pip install knowledge-rag knowledge-rag init # scaffolds config.yaml + documents/ Drop your PDFs, markdown, code files into documents/. Restart your MCP client. Ask it: search_knowledge('your query'). First query downloads the ONNX embedding model (~200MB). Subsequent queries are cached.

License and pricing

MIT license. Open source. No paid features.

Based on the project's public README and repository metadata.

Add this badge to your README

Featured on YunoDB Badge styles & help

Maintain this project? Show that it is featured on YunoDB.

Markdown[![Featured on YunoDB](https://yunodb.com/badge/lyonzinknowledge-rag.svg)](https://yunodb.com/tools/lyonzinknowledge-rag?utm_source=github&utm_medium=badge)
HTML<a href="https://yunodb.com/tools/lyonzinknowledge-rag?utm_source=github&utm_medium=badge"><img src="https://yunodb.com/badge/lyonzinknowledge-rag.svg" alt="Featured on YunoDB" height="20"></a>