Made for the way you work
Follow-ups and smart waits
Keep the next step attached to the work. Resume when there is something useful to do.
In practice
A useful place to start
Ask the agent to check a build when it finishes, then inspect the output. Or schedule a follow-up in the same chat for a specific time.
Wake on a result
A chat can wait for a background job, process, file change, HTTP result, or CI completion. The wake-up includes evidence that the agent can inspect.
Set a later task
One-off scheduled tasks keep instructions in the current chat and start a follow-up when due. They are distinct from loops, which define recurring work.
Keep pending work visible
See pending waits and tasks, cancel work that is no longer needed, and inspect whether a wait completed, failed, or expired. Stored waits can survive an editor restart.
Get started
Tell the agent what event or time matters and what to do once it happens. Use an explicit timezone for a scheduled time.
Good to know
One-off tasks are limited to the supported scheduling window, currently up to 30 days. The host must be available to run work; failed probes and timeouts are not proof that the intended result happened.