Home / Features / Agents / Schedule agents on cron
Schedule AI agents on a cron with osFoundry
osFoundry schedules any agent profile to run on a cron — every minute, hour, or day. Wrap a profile with a trigger and a starting prompt; the agent runs unattended, streams logs you can replay later, and pings you when something needs your attention. Pay only for the seconds the agent actually runs.
Quick answer
- Cron triggers on any agent profile (every minute / hour / day / custom expression).
- Other triggers: webhook, channel event, channel message.
- Live log streaming + replay.
- Pay only for seconds of actual compute.
Key capabilities
- Standard cron expressions, down to per-minute granularity.
- Trigger types: cron, webhook, channel event, manual.
- Per-run log streaming with replay.
- Configurable retry + backoff on failure.
- Ping flows — Slack, email, in-app notification, custom webhook.
- Per-run cost meter visible in the run history.
How to do it in osFoundry
- Define an agent profile — In osStudio → Agents, create the profile: prompt, tools, memory groups, model.
- Open Automations — In /agents → Automations, click "New automation".
- Pick a trigger — Cron (`*/5 * * * *` for every 5 minutes), webhook, or channel event. Provide the starting prompt the agent receives on each tick.
- Add a notification step — Optional but recommended — wire a ping step at the end so you know when the agent finds something important.
Use cases
- Sales ops: Every morning at 7 AM, an agent scans yesterday’s pipeline changes, summarises deal movement, posts to a Slack channel.
- Brand monitor: Every hour, an agent checks Hacker News and Reddit for mentions of your product, pings you only when sentiment shifts.
- Cost-sensitive ops: Nightly bill-anomaly agent reads your cloud invoices, flags >20% week-over-week spikes, opens a ticket.
Frequently asked questions
How granular can the schedule be?
Standard cron expressions — down to per-minute granularity.
What happens if a scheduled run fails?
It is logged with the failure, optionally pings you, and retries on the next scheduled tick (configurable backoff).
Are scheduled agents billed when idle?
No — only when actively running.
Can I pause a schedule without deleting it?
Yes — toggle "active" off. The schedule stops firing; resume by toggling back on.
Can two automations share an agent profile?
Yes — profiles are reusable. Different triggers, same agent definition.
Pricing
Per-second of agent runtime + per-token of model usage (BYOK to your provider). Idle schedules cost $0.
Related features