MCP integration

[Yeon1007]

Does RAGFlow currently integrate any MCP(Model Context Protocol) related functions or components?

[dosubot[bot]]

Yes, RAGFlow currently integrates MCP (Model Context Protocol) related functions and components. It includes a dedicated MCP server that can be launched alongside the main RAGFlow server, supporting both self-host and host modes with API key authentication. The MCP server enables deep document retrieval and tool invocation via HTTP+SSE or streamable HTTP transports.

There are comprehensive API endpoints for managing MCP servers and tools (list, create, update, delete, import, export, test), and the UI provides dialogs, forms, and cards for MCP operations in user settings. Key implementation files include mcp/server/server.py, api/apps/mcp_server_app.py, and common/mcp_tool_call_conn.py—these handle server logic, API endpoints, and client connections respectively. Official documentation covers setup, launch, and client usage examples, and recent pull requests confirm active development and ongoing improvements to MCP integration. For more details, see the MCP server documentation and code references citation 1 citation 2 citation 3.

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}