Scheduled runs
Trigger a saved RunRequest on a cron-style cadence so detection coverage is validated continuously instead of ad-hoc. Pro plan and above.
Key points
- Cadence is a discriminated union: every_hours, daily, or weekly. The scheduler loop ticks every 60 seconds.
- Each schedule wraps a frozen RunRequest template; mutating it requires editing the schedule directly.
- Skipped runs are recorded — they do not silently disappear from the audit chain.
- Disabled schedules retain history; re-enable to resume from the next cadence tick.
01
When to use one
- Nightly regression on a production-shape staging endpoint.
- Hourly canary against a high-traffic chatbot to catch silent regressions.
- Weekly compliance posture sweep for SOC 2 / NIST AI RMF evidence.
02
Operational notes
- Schedules respect the runs/month quota — they do not bypass it.
- A long-running deep+high+agentic run will still complete past its scheduled window; the next tick fires only after the current run terminates.
- Pair with webhooks for a hands-off feedback loop into Slack or PagerDuty.