Automations & scheduled tasks
Recurring AI jobs: "run tests every morning," "deploy every Friday," "summarize new issues each hour."
What an automation is
An automation is a saved prompt that runs on a schedule in a chosen workspace with a chosen model. Each field:
| Field | Meaning |
|---|---|
| Name | Label in the automations list. |
| Prompt | What the agent should do each run. |
| Model | Which model runs it. |
| Workspace | Where it runs (full tool access in that workspace). |
| Schedule (RRULE) | An iCalendar recurrence rule — e.g. every day at 09:00, every weekday, first of the month. AI Labs shows the next few run times. |
| Active | Toggle on/off without deleting. |
Runs are recorded with their output, and a new chat from an automation shows up on the landing page.
Webhook triggers
Each automation can also be run on demand via a webhook:
- Generate a webhook token for the automation — AI Labs returns a one-time URL of the form
.../api/automations/<id>/run?token=<token>. POSTto that URL to trigger a run. Any JSON body is injected into the prompt as context, so an external system can pass a payload (a CI result, an alert, a form submission).- Regenerate or revoke the token at any time. Tokens are stored hashed.
Managing automations
The Automations panel (also reachable as Scheduled) lists every automation with its next run, last run, and status. The agent can also manage automations from chat with the list_automations / create_automation / update_automation / toggle_automation / delete_automation tools.
The scheduler is a supervised background worker — if it fails it restarts with backoff, and its health shows in the detailed health check. The sidebar link is hidden when chat is disabled.