datahub-semantic-agent 0.1.3

MCP server for talking to your data, powered by DataHub

datahub-semantic-agent

An MCP server for talking to your data, powered by DataHub.

Usage

The datahub-semantic-agent requires a number of environment variables.

# DataHub connection details.
DATAHUB_GMS_URL=...
DATAHUB_GMS_TOKEN=...

# OpenAI API key.
# Required for running our internal agentic workflows.
# We don't currently support any other LLM providers.
OPENAI_API_KEY=sk-...

# Snowflake connection details.
# The options mirror the configs available in the DataHub Snowflake integration.
# https://docs.datahub.com/docs/generated/ingestion/sources/snowflake/
# e.g. stuff like this is also supported:
#    SNOWFLAKE__CONNECTION__AUTHENTICATION_TYPE: KEY_PAIR_AUTHENTICATOR
#    SNOWFLAKE__CONNECTION__PRIVATE_KEY: ...
#    SNOWFLAKE__CONNECTION__PRIVATE_KEY_PASSWORD: ...
SNOWFLAKE__CONNECTION__ACCOUNT_ID=...
SNOWFLAKE__CONNECTION__USERNAME=...
SNOWFLAKE__CONNECTION__PASSWORD=...
SNOWFLAKE__CONNECTION__WAREHOUSE=...
SNOWFLAKE__CONNECTION__ROLE=...

Then, configure your MCP client (Claude Desktop, Cursor, Windsurf, etc.) with the following settings:

{
  "mcpServers": {
    "datahub": {
      "command": "<path-to-uvx>",  // e.g. "/Users/hsheth/.local/bin/uvx"
      "args": [
        "mcp-server-datahub",
      ],
      "env": {
        "DATAHUB_GMS_URL": "...",
        "DATAHUB_GMS_TOKEN": "...",
      }
    },
    "datahub-semantic-agent": {
      "command": "<path-to-uvx>",  // e.g. "/Users/hsheth/.local/bin/uvx"
      "args": [
        "datahub-semantic-agent",
      ],
      "env": {
        "DATAHUB_GMS_URL": "...",
        "DATAHUB_GMS_TOKEN": "...",
        "OPENAI_API_KEY": "sk-...",
        "SNOWFLAKE__CONNECTION__ACCOUNT_ID": "...",
        ...
      }
    }
  }
}

The DataHub MCP server must also be configured for the datahub-semantic-agent MCP server to work properly.