RAG_MCP

A RAG-ready MCP server for semantic PDF search with OCR, FAISS, and transformers—plug into any MCP client and retrieve intelligent answers within your MCP client.

Step 1: Create virtual env and install requirements

curl -LsSf https://astral.sh/uv/install.sh | sh
uv init rag_mcp
cd rag_mcp
uv venv
source .venv/bin/activate
uv pip install -r requirements.txt
brew install tesseract

Step 2a: Add config to the Claude MCP client

code ~/Library/Application Support/Claude/claude_desktop_config.json
...
{
    "mcpServers": {
        "rag": {
            "command": "/Users/XXX/.local/bin/uv",
            "args": [
                "--directory",
                "/Users/XXX/Documents/RAG_MCP",
                "run",
                "rag.py"
            ]
        }
    }
}

Step 2b: Add config to the Cursor MCP client

code ~/.cursor/mcp.json
...
{
    "mcpServers": {
        "rag": {
            "command": "/Users/XXX/Documents/RAG_MCP/start.sh",
            "args": []
        }
    }
}

Step 5: Make MCP server executable

chmod +x start.sh
chmod +x rag.py

Step 6: Run MCP server (Claude Desktop)

uv run rag.py

Step 7: Run MCP client and query

Parse the pdfs and tell me about 18 Church St. and what significance it has.

License

Apache License, Version 2.0