Skip to main content

​
Overview

The datafrey-mcp package is the MCP server that https://mcp.datafrey.ai/mcp runs. It’s Apache-2.0 and you can run it yourself — locally or on your own infrastructure.
Self-hosting is optional. The hosted endpoint at https://mcp.datafrey.ai/mcp works out of the box. Self-host if you want.
Self-hosting only replaces the MCP bridge. Queries still go through the hosted DataFrey API, so you still need a DataFrey account and a connected database.

​
Install

From source (uv workspace): 
git clone https://github.com/datafrey-ai/datafrey
cd datafrey
uv sync --package datafrey-mcp
uv run datafrey-mcp
Or via Docker: 
docker build -f packages/datafrey-mcp/Dockerfile -t datafrey-mcp .
docker run -p 8080:8080 datafrey-mcp
The server listens on 0.0.0.0:8080 by default over streamable HTTP.

​
Point your client at it

Instead of https://mcp.datafrey.ai/mcp, use your local URL: 
{
  "mcpServers": {
    "datafrey": {
      "url": "http://localhost:8080/mcp"
    }
  }
}
The OAuth flow and the plan / run tools behave the same as the hosted version.

​
Source