Release
Agents API is now generally available
Build multi-step agents that call your own tools, with retries and logs built in.
Agents are now available to every workspace. An agent can plan a task, call the tools you register, check its own work and hand back a result.
Every step is logged, so you can replay a run, see which tool was called and why, and fix a prompt without guessing.
Start with one of the example agents in the dashboard, or register your first tool in a few lines of code.
