Use the Databricks Supervisor API

The Supervisor API lets Databricks run the tool-selection and synthesis loop server-side. Instead of your agent managing tool calls and looping, you declare hosted tools and call responses.create() — Databricks handles the rest. You can also declare client-side function tools (Python callables your app executes) alongside hosted tools in the same request.

When to Use

Use the Supervisor API when you want to:

• Connect Genie spaces, UC functions, Knowledge Assistants, or MCP servers without managing the agent loop yourself
• Mix client-side function tools (your own Python callables) with hosted tools in the same request
• Choose models at runtime and control which tools are used per request
• Offload tool orchestration to Databricks while iterating on your agent

Limitations:

• Inference parameters (e.g., temperature, top_p) are not supported when tools are passed
• stream and background cannot both be true in the same request
• Background mode requests have a maximum execution time of 30 minutes

Step 1: Install databricks-openai