Connect agents to OpenRAG knowledge bases over MCP 是做什麼的？
Use OpenRAG's built-in MCP endpoint so agents can ingest documents, search a knowledge base, create filters, and run RAG-backed chat against a deployed OpenRAG instance.
Prerequisites
OpenRAG deployment; OpenRAG API key; MCP client with streamable HTTP support such as Cursor, Claude Desktop, IBM Bob, or another MCP-capable agent runtime
Installation
Deploy or run OpenRAG first, then connect your MCP client to the built-in HTTP endpoint.
For the documented self-managed deployment, install Python, uv, and Docker or Podman, then clone and configure the upstream project:
- git clone https://github.com/langflow-ai/openrag.git
- cd openrag
- cp .env.example .env
Configure the required .env values such as OPENSEARCH_PASSWORD, model provider credentials, and Langflow credentials. Then start Docling with the documented uv command, using scripts/docling_ctl.py start --port 5001, before launching the OpenRAG containers:
- docker compose up -d
For a terminal-managed evaluation install, upstream also documents uvx --python 3.13 openrag.
Configure your MCP client to use the running OpenRAG instance at /mcp with the same OpenRAG API key used for the REST API:
{ "mcpServers": { "openrag": { "url": "http://localhost:3000/mcp", "headers": { "X-API-Key": "orag_your_api_key_here" } } } }
Do not install the deprecated openrag-mcp package; OpenRAG now serves MCP directly from the /mcp endpoint.
- Source: https://github.com/langflow-ai/openrag
- Extracted from upstream docs: https://raw.githubusercontent.com/langflow-ai/openrag/HEAD/README.md