Skip to main content
An agent can run on its own on a schedule. Set a cadence and the scheduler triggers background runs for you — no one has to be at the dashboard. Each scheduled run is recorded just like any other run, with its own output chat.

​
Schedule modes

​
manual
default
No schedule. The agent runs only when you start it interactively or through the API. This is the default.
​
daily
cadence
Runs once a day at a time you choose.
​
weekly
cadence
Runs once a week on the day and time you choose.
​
monthly
cadence
Runs on a monthly cadence — including every N months — at the time you choose.
​
custom
cron expression
Any standard cron expression, for full control over the cadence.
You can enable or disable a schedule at any time without redeploying the agent, so pausing and resuming automated runs is instant.
For a repeat like “every N days” or “every N weeks” (any interval other than every 1), use a custom cron expression that spells out the exact days you want. Cron gives you precise, predictable cadences for these patterns.

​
Schedules run in UTC

All schedules run in UTC. Set your run time with UTC in mind.
If your local time observes daylight saving, your agent’s local run time will shift by an hour when the clocks change, because the schedule itself stays anchored to UTC. Choose a time that’s acceptable year-round, or adjust the schedule seasonally when the clocks change.

​
What a scheduled run needs

A couple of things need to be in place for scheduled runs to fire successfully. The dashboard shows whether scheduled runs are supported for your agent, so you can confirm at a glance.
Scheduled runs consume credits like any other run. If the tenant is out of credits, the run is skipped and marked accordingly rather than failing silently — so your credit balance is always the single thing to top up.
If the agent uses connected apps, it needs a bound identity so those apps can act on someone’s behalf during an unattended run. Agents driven by an API key need a bound subject for this. Without one, scheduled runs can’t use connected apps — the dashboard indicates whether scheduled runs are supported so you know before you rely on them.

​
How scheduled runs are recorded

Each scheduled run is recorded as a run with its own output chat, so your automated work has the same history and reviewability as interactive runs. See Running & run history.
Retries never double-run the same scheduled occurrence — a given occurrence produces exactly one run, so automated work stays clean and idempotent.

​
Next steps

Review scheduled runs

Open the chats your scheduled runs produce.

Configure connected apps

Set up the integrations a scheduled agent will use.