Connect Genie Code to Prometheux using the Remote MCP Server. Genie Code is Databricks’ autonomous AI agent for data work — it supports MCP as an extensibility mechanism, allowing it to securely interact with external tools and APIs. By registering Prometheux as an external MCP server, Genie Code gains the ability to list concepts, run reasoning, and query knowledge graphs — directly within the Databricks environment. Databricks registers external MCP servers as MCP services: Unity Catalog objects addressed as <catalog>.<schema>.<name>, each backed by an HTTP connection that holds the server URL and OAuth settings. See the Databricks guide to registering external MCP servers.

Prerequisites

  • Databricks workspace with Unity Catalog and access to Genie Code
  • Unity Catalog permissions on the schema you register into: USE CATALOG on the catalog, USE SCHEMA, CREATE SERVICE and CREATE CONNECTION on the schema
  • Prometheux credentials (username, organization, API token)

Setup

In your Databricks workspace, open MCPs, click + MCP, and choose Connect an existing MCP server.

Step 1 — Name

Step 2 — Connection

Click Create new connection and fill in: The form asks you to add https://<your-workspace>/login/oauth/http.html to your OAuth app’s allowed redirect URLs. No action is needed: the Prometheux server already accepts redirects to any Databricks workspace domain.
The Server URL must include the /px-remote-mcp-server/mcp path. With only https://api.prometheux.ai, tool loading fails with a 405 Not Allowed HTML response.

Step 3 — Sign in and load tools

  1. Click Create & load tools. This saves the connection.
  2. Under Tools, click Login. In the browser tab that opens with the Prometheux login form, enter:
    • Username: Your Prometheux username
    • Organization: Your Prometheux organization
    • API Token: Your Prometheux authentication token
  3. Submit the form — Databricks receives your session token and lists the Prometheux tools.
  4. Choose the tools to expose, or leave all of them selected.

Step 4 — Create the MCP service

Click Create MCP Service.
Loading tools in Step 3 only saves the connection. The MCP is not registered — and does not appear in the MCPs list or in Genie Code — until you click Create MCP Service.

Usage

Once the MCP service is created, Genie Code can use the Prometheux MCP tools in your agent workflows. Other agents refer to it by its full name, <catalog>.<schema>.<name>; clients that need a URL use:
To let other users invoke it, grant them EXECUTE on the MCP service. They do not need USE CONNECTION on the underlying connection, but each user signs in with their own Prometheux credentials.
Available ToolsSee the MCP overview page for the complete list of tools exposed to Genie Code.

Session expiry

The connection stays signed in until your Prometheux API token expires — Databricks shows this as the connection’s Access token expiration. After that, create a new API token and click Login on the connection again.

Troubleshooting

”405 Not Allowed” or “unexpected non-JSON (HTML) response”

The connection’s Server URL is missing the server path, so requests reach the Prometheux front proxy instead of the MCP server. Solution:
  • Edit the connection and set Server URL to https://api.prometheux.ai/px-remote-mcp-server/mcp
  • Click Login again and reload the tools

MCP does not appear in the MCPs list

The connection was created but the MCP service was not. In Catalog Explorer, the schema shows the connection under Connections but Services is 0. Solution:
  • Start again from + MCP > Connect an existing MCP server, choose the same catalog and schema, select the existing connection, and click Create MCP Service
  • If creation fails, ask a workspace admin for CREATE SERVICE on the schema, or register into a schema where you already have it

Login tab doesn’t open

Solution:
  • Check your browser’s popup blocker settings for the Databricks domain
  • Try the connection flow in an incognito/private window

Tools not appearing in Genie Code

Solution:
  • Ensure the login step completed successfully and the MCP service exists in the MCPs list
  • Check that the tools you need were selected when creating the MCP service
  • Verify your Prometheux credentials are valid
For general troubleshooting, see the Remote MCP Server troubleshooting section.