mcp-server-datahub

A Model Context Protocol server implementation for DataHub. This enables AI agents to query DataHub for metadata and context about your data ecosystem.

Supports both DataHub OSS and DataHub Cloud.

Features

• Searching across all entity types and using arbitrary filters
• Fetching metadata for any entity
• Traversing the lineage graph, both upstream and downstream
• Listing SQL queries associated with a dataset

Usage

uv sync --no-sources

For authentication, you can either use datahub init to configure a global ~/.datahubenv file, or you can set the appropriate environment variables:

datahub init   # follow the prompts

# Alternatively, use these environment variables:
export DATAHUB_GMS_URL=https://name.acryl.io/gms
export DATAHUB_GMS_TOKEN=<your-token>

Claude Desktop

mcp install mcp_server.py

Cursor

# In .cursor/mcp.json
{
  "mcpServers": {
    "datahub": {
      "command": "<path>/mcp-server-datahub/.venv/bin/mcp",
      "args": [
        "run",
        "<path>/mcp-server-datahub/mcp_server.py"
      ],
      "env": {}
    }
  }
}

Other MCP Clients

command: <path>/mcp-server-datahub/.venv/bin/mcp
args:
  - run
  - <path>/mcp-server-datahub/mcp_server.py

Developing

Setup

uv sync --no-sources
# Alternatively, if also developing on acryl-datahub:
# Assumes the datahub repo is checked out at ../datahub
uv sync

# <authentication is the same as above>

Run using the MCP inspector

source .venv/bin/activate
mcp dev mcp_server.py

Run tests

The test suite is currently very simplistic, and requires a live DataHub instance.

pytest