# Schedules

## How does a notebook run on its own?

A notebook can run on its own. Turn on **Run automatically** under **Settings → General** on the
notebook, then pick an interval or a time of day. Each scheduled run starts from a fresh kernel, so
its outputs always match the code.

A scheduled run runs as whoever turned the schedule on, on their data access and their compute,
whoever owns the notebook. It runs the notebook as it stands, including cells someone else changed
since. Turning a schedule on takes edit access, and the next person to turn it on becomes the person
it runs as.

Your agent sets a schedule with `save_notebook`, passing the `notebookId` and a `cadence`. The
built-in assistant asks on a confirmation card before it turns one on, since a schedule spends your
compute on a timer.

## What happens when a run fails?

- The first failed run after a good one notifies the person it runs as and the owner.
- Three failed runs in a row, spread across at least a day, pause the schedule until someone turns it
  back on.
- When that person's compute allowance runs out, the schedule waits and resumes on its own when the
  allowance resets.
- A notebook that no member has opened, and nobody has read through its public view, for 30 days stops
  running. A member opening it starts the schedule again.