What it does

Triggers start an agent from an event instead of a user message. A record is created, an email arrives, a schedule ticks over — the trigger fires and the agent runs. They're what turn an agent autonomous.

Key facts

• Triggers are built on Power Automate connectors, so anything with a connector trigger can start an agent.
• Event data passes in as inputs the agent's instructions and tools can use.
• Multiple triggers can be attached to one agent for different entry points.
• Each trigger runs under a connection, and that connection's identity governs what the run can access.

When to use / skip

Use triggers when work should start without anyone asking — proactive notifications, automatic triage, scheduled housekeeping. Skip them where a human should always initiate; an event-driven agent that fires unexpectedly erodes trust quickly.

Configuration decisions

• Which events justify an autonomous run vs a user-initiated conversation.
• The connection and identity each trigger uses.
• Rate and volume limits so a burst of events doesn't overwhelm capacity.

Gotchas

• Chatty source systems fire far more often than people expect — a noisy trigger can drain capacity overnight.
• Trigger connections silently break when credentials expire; the agent just stops firing. Monitor them.

Consultant notes

• Design triggers with idempotency in mind — the same event may arrive twice.
• Keep trigger connections on managed service identities, not a maker's personal account.
• Build alerting on trigger failures; silent stops are the common production incident.

---

Recheck supported trigger sources each release.