ragflow

RAGFlow MCP Server exposes APIs to search knowledge bases and engage in chat-based interactions. It provides three primary tools: list_datasets (lists available datasets with IDs and names), create_chat (creates a new chat assistant given a name and a dataset_id, returning the chat_id, name, and session_id), and chat (conducts a conversation with a chat assistant using a session_id and a user question, returning the assistant’s reply). To run the server, start it via the UVX command shown in the configuration, which wires the server entrypoint ragflow-mcp-server with the required API key and base URL. Once running, you can discover datasets, spin up chat assistants tied to specific datasets, and then have interactive conversations through the chat endpoint. For debugging and inspection, you can use the MCP Inspector tool, which can attach to the server process and provide a guided debugging UI.

Prerequisites:

• Python 3.8+ (or a compatible Python environment)
• Access to install UV or UVX utilities as described below

Install and run (example using UV/UVX workflow):

1. Install the UV/UVX tooling if you don’t have it yet. This typically involves installing uv/uvx via your preferred method (e.g., pipx or a package manager).
2. Ensure you have a Python environment and Python dependencies installed for the ragflow-mcp-server.
3. Start the MCP server using the configuration shown in the repository (uvx ragflow-mcp-server --api-key=... --base-url=...).

Optional tooling:

• MCP Inspector for debugging: install via npm and run the inspector command shown in the README to attach to your running server.

Notes:

• The base URL and API key are required to access the RagFlow MCP endpoints; keep credentials secure.
• If you update dependencies, regenerate or verify dist artifacts if your setup requires them.

Tips and considerations:

• The server uses standard IO for its interfaces, which can complicate debugging. Leverage the MCP Inspector for a structured debugging experience.
• Ensure the base URL (e.g., http://172.16.33.66:8060) is reachable from the environment where you run the MCP server.
• If you encounter authentication or connectivity issues, verify the API key and network access to the RagFlow services.
• The configuration examples show both development/unpublished and published server settings; adapt the command and arguments to match your deployment scenario.