Triggers
How an agent starts: manually, on a schedule, from a webhook, when an alert fires, or on a Slack event.
A trigger decides when an agent runs and hands it the first input. Every agent has exactly one. The trigger itself is never billed; see Plans and billing.
Manual
Runs when you press Run in the editor, or when another agent calls it. Use it for testing and for reusable building blocks.
Schedule
Runs on a schedule in the timezone you choose, for example every Monday at 09:00. Use it for digests, nightly reports and periodic checks.
Webhook
Runs when something POSTs to the agent's signed URL. The request is verified with a secret, can require an HMAC signature header, and de-dupes on an idempotency key. Use it to start an agent from an external system.
Alert
Runs when any of the chosen alerts fires, with that alert and its value as the input ({{ $trigger.alert }}). Pick one alert or several: the agent runs once per firing, whichever one fired.
This is how thresholds become action. An alert decides when; the agent decides what to do: have Lucid investigate the failures into issues, post to Slack, hand an issue to a coding agent, or improve an eval. Track issues deploys an alert and its triage agent together.
Slack events
Slack can start an agent too: a new message in a channel, an @-mention of the Trodo bot, a reaction, a shared file, a new channel or a new user. See Slack.