Genie Agent MCP server
The Genie Agent MCP server lets an agent query a single Genie Agent with natural language, without writing SQL. The agent sends a plain-language question, and Genie writes and runs the SQL against that agent's tables and returns the result, with Unity Catalog permissions enforced on every request.
To query a Genie Agent from Python without MCP, use AI Bridge tool examples. For business questions across your workspace, use the Genie One MCP (system.ai.genie_one_mcp).
A Genie Agent is a collection of up to 50 Unity Catalog tables, views, or metric views that Genie keeps in context, so the server is scoped to that curated data rather than your whole workspace.
URL pattern | OAuth scope |
|---|---|
|
|
To find the Genie space ID, open the Genie Agent and copy the space ID from its Settings tab.
Calls to this endpoint don't go through Unity Gateway, so access is controlled by permissions on the Genie Agent and its underlying data.
Genie Agent MCP calls use Databricks SQL pricing.
Prerequisites
A workspace administrator must enable Managed MCP Servers on the workspace Previews page.
Connect an agent
To connect an agent to a Genie Agent MCP server, create a Genie Agent, share it with the users or service principals that need access, and point your agent at the pre-configured MCP URL for the space. Follow Use MCP tools in a Python agent, setting server_url to the URL above.
To connect external clients such as Cursor or Claude Desktop, see Supported coding agents.
Limitations
- Read-only. The Genie Agent MCP server queries data; it doesn't write to your tables.
- No conversation history. The server invokes Genie as a tool, so it doesn't pass conversation history to the Genie API. To continue a conversation from Python, use AI Bridge.