Schedules and triggers

Run work on a timetable, or the moment an event fires, and let Workspace Admins pause or start it under System settings › Schedules.

Copy page for AIUpdated November 8, 2026

Some work should just happen: a reminder every morning at 7, a sync every Friday, a check whenever a new record arrives. Schedules run your module's jobs on a timetable. Triggers run your agent the moment an event fires. Both are declared in your module and show up in every workspace that installs it.

Schedules: every morning at 7, every Friday

A schedule names a job in your module and when to run it, written as a cron expression with a timezone:

You wantCron
Every day at 7:000 7 * * *
Every Friday at 16:000 16 * * 5
Every 15 minutes*/15 * * * *

You choose whether it starts on when a workspace installs your module, or is offered for a Workspace Admin to switch on. Offering is right for anything that touches the outside world, like reading a mailbox.

A scheduled job has no person behind it, so it does bookkeeping your module does for itself: sending reminders, tidying up, copying data in.

Triggers: when a record is created, run the agent

A trigger joins an event to your agent, with a short instruction: "Triage job {payload.id}." When the event fires, the agent runs.

There's no person present, so the agent acts as its own identity, which holds only the roles a Workspace Admin gives it. With no role, the run is refused. See Agents and AI models.

Turning them on and off under System settings › Schedules

A Workspace Admin manages them in one place:

  1. Open System settings, then Schedules.
  2. Click a schedule to see when it runs, its timezone and its History.
  3. Click Pause to stop it, or Resume to start it again. Run now runs it once, straight away.
  4. Under Offered by modules, click Turn on to start a job a module offers. It runs as the admin who turned it on.

Module-managed shows only the schedules modules created. Fire event sends a test event, to check a trigger without waiting for the real thing.

Keeping a job safe to run twice

Jobs can run twice: a retry after a hiccup, or Run now pressed on a busy day. Make every job safe to repeat:

  • Check before you act. "Already reminded about this due date?" Skip it.
  • Use a dedupe key for notifications, so a person is told once.
  • Start long work and return. A job gets 60 seconds per call. Start the slow part (a long Bridge job or a compute job) and pick up the result next time.

Ask the Module Builder

For developers

json
"scheduler": {
  "entity_types": [{"entity_code": "DAILY_REMINDERS", "name": "Daily reminders",
    "handler": "JOBS.send_reminders", "default_cron": "0 7 * * 1-5",
    "timezone": "America/Los_Angeles", "auto_create": true}],
  "agent_triggers": [{"trigger_code": "TRIAGE_NEW", "name": "Triage new jobs",
    "event": "jobs.job_created", "agent_code": "JOBS_AGENT",
    "prompt": "Triage job {payload.id}.", "auto_create": true}]
}

The handler is a service your module registers (and declares) with no permission, because a schedule has no user.

Good to know

  • Schedules are included; they aren't metered. What the job does may be, such as AI use.
  • Pausing or deleting keeps the history, so you can still see what ran.
  • Work on outside input (an email, a web form) should start the agent with only the tools that job needs, and treat the content as data, never as instructions.

Can't find what you need? Contact support.