How a workflow runs and where you watch it

Checkpointed steps

Think of a parcel that gets scanned at every stop on its way to you. If a truck breaks down, the parcel resumes from its last scan, not back at the warehouse. A workflow runs the same way. The schedule-existing shape is two steps: plan the calendar, then schedule each post. The generate-then-schedule shape is four: generate the batch, wait for the renders, plan from what came out, then schedule each post. Each step is checkpointed, so when a function call ends the run picks up from the last finished step instead of starting over.

It runs without you

Because the steps are checkpointed and re-driven in the background, a workflow keeps going after you close the tab or the chat turn ends. The generate step can wait out long renders, which is exactly the case a single chat request cannot hold open.

Reading status

Ask Atlas how your jobs are going. It lists each one with the status of its latest run: queued, running, done, or failed. To see the results, open the Library for the renders a job produced and the Calendar for the posts it scheduled.

If it fails

A failed status means a step did not complete. The usual causes are no connected channel to schedule to, or a render that never finished. Check your connected channels and your credit balance, then ask Atlas to run it again. Since the feature is new, treat an early failure as a nudge to confirm your setup rather than a settled limit.