概览
What is 🚀 R2R MCP Server?
R2R MCP Server provides a Model Context Protocol (MCP) integration with the R2R (Retrieval-to-Response) system, enabling Claude and other MCP-compatible models to access knowledge bases through tools like search, RAG, web search, and agent research.
How to use 🚀 R2R MCP Server?
Install Python 3.12+, uv, clone the repository, create a virtual environment, install dependencies via make install or uv pip install ., then run mcp install app/server.py -v R2R_API_URL=... or -v R2R_API_KEY=... to register as an MCP plug-in. The server exposes tools such as search, rag, web_search, document_search, list_documents, and agent_research. Use make run to start locally.
Key features of 🚀 R2R MCP Server
- MCP server for Claude and other MCP-compatible models
- Asynchronous R2R client for high-performance knowledge base access
- Search and RAG tools with vector, full‑text, and hybrid search
- Structured logging via Loguru
- Lightweight architecture with minimal dependencies
- Web search and agent research for complex queries
Use cases of 🚀 R2R MCP Server
- Search documents and retrieve specific text fragments
- Answer questions with citations from your knowledge base
- Conduct multi‑source research by combining different information sources
- Fetch up‑to‑date information from the web when needed
FAQ from 🚀 R2R MCP Server
What tools does the server expose?
It provides search (vector/hybrid), rag (answer with citations), web_search, document_search, list_documents, and agent_research for complex LLM‑driven research.
What are the runtime dependencies?
Python ≥3.12, the uv package manager, and three core libraries: mcp, r2r, and loguru. Development and testing require additional packages like pytest, black, mypy, etc.
How do I configure the R2R API connection?
Set the R2R_API_URL (e.g., https://api.sciphi.ai) or R2R_API_KEY environment variable when installing the server with mcp install app/server.py -v.
Can the server work with models other than Claude?
Yes—it is compatible with any LLM that supports the Model Context Protocol, including custom MCP clients.
Are there performance benchmarks?
Yes, the project includes dedicated performance tests in test_r2r_mcp_performance.py that measure formatting speed and tool processing throughput.
