Trigger

In practice

Triggers are what separate an agent you use from an agent that works. A scheduled trigger fires at a fixed cadence, which suits recurring deliverables such as a daily digest or a weekly report. An event trigger fires when something happens, which suits reactive work such as triaging a message as it lands. The detail that catches people is that a schedule on a machine that is asleep does not queue up and run later; the moment passes and nothing records that it did. Recording every trigger, including runs that found nothing to do, is what makes silence diagnostic.

Example

A recruiter uses two triggers on one agent. A scheduled trigger runs at 8am every weekday and summarises new applications. An event trigger fires whenever a candidate books an interview, through the booking tool's webhook, and the agent prepares a short brief on that candidate. Every trigger leaves a run record, including mornings with no new applications, so a quiet Monday looks different from a broken one.

Common questions about trigger

Read more

• Schedules and triggers
• AI agent monitoring

Nearby terms

• Spend cap
• Unattended execution
• White-label AI agents