Scheduled jobs

Recurring work against your systems and cloud accounts. Create jobs from Garry: ask for something recurring (“every night at 2am, check disk space on #prod”) and Garry shows the exact command or goal and the schedule for you to approve.
| Kind | Runs |
|---|---|
| Command | One fixed command on the target systems, exactly as approved |
| Agent | A goal (“summarize yesterday’s AWS bill”). The AI works it out at each run, using read-only commands and cloud calls only: nobody is there to approve a change |
Each job shows its schedule in words and as cron (schedules are in UTC unless the job names a time zone), its targets (systems, tags or a cloud account), the next and last run and their status.
| Action | |
|---|---|
| Run now | Run at once, without changing the schedule |
| Pause | Stop running until resumed |
| Edit | Change the name, schedule or targets |
| History | Past runs with their output |
| Delete | Remove the job |
Targets are resolved at each run, so a job on #prod includes systems tagged later. A run that
finds no connected target is recorded as skipped, not failed. Agent jobs notify their owner
when they finish.