This section covers which AI models power an agent's different functions, and which ready-made, built-in abilities you switch on for it.
Overview
Models is a set of pickers, grouped by stage (Task Processing, Indexing, Pre-Processing, Response Generation, Guardrails), that decide which model does what for the agent. Every function must always have a model, and the available choices come from your workspace's Model Library catalog. Capabilities is a different thing entirely: a settings list of the platform's built-in functions — web search, running code, generating a chart or document — that you simply switch on, with no connection to set up. Together they answer "which brain does the thinking" and "what can it already do out of the box."
What's in This Section
Topic | Covers |
|---|---|
Models | Assigning a model to each agent function, and where inference parameters like temperature live instead. |
Capabilities | Turning built-in functions on and off, and how Capabilities differs from Tools and Skills. |
Capabilities vs. Tools
If you're looking for a built-in ability the agent already has, check Capabilities first — the standard tools moved here. Reach for Tools & Skills (a separate section) only when you need to connect the agent to your own systems: a connector action, an MCP server, or a custom endpoint.