Background runs that report back
Hand the agent a long job, such as research across many sources, and keep chatting while it works in the background.
What it includes
Watch it work
A card in the thread shows whether the run is waiting, running, finished, failed or canceled, and opens a page that fills in live.
The answer lands in your chat
The finished result reaches the agent in the chat that started it, so the conversation picks up where the work left off.
Check on it or stop it
Ask how the run is going or ask to stop it, and the agent checks or cancels it for you.
Full transcript for every run
Open any scheduled, webhook or delegated run to read everything the agent did, step by step, with how it ended and how long it took.