AI Agents

In one line

AI that doesn't just answer but takes steps: searching, using tools, and checking its own work toward a goal.

Picture it

Going from a helpful librarian to an assistant who actually runs the errands.

Under the hood

A model in a loop that your code runs: it receives the context, either answers or proposes a tool call, your code validates and runs permitted calls and returns the results, and the loop continues until the goal is verified or a limit, denial, cancellation or handoff ends it. The model's stop reason says why one response ended, not whether the task succeeded.