Trigger Hub and notifications
Trigger Hub turns a proven Agent Team into a guarded automation. One primary trigger starts each deployment, while Projects keeps health, tests, deliveries, runs, and cost visible.
Choose one start event
Open Automate from the Agent Team toolbar or Automate this proof from a completed personalized proof, then choose Scheduled, App event, or Webhook. Scheduled supports hourly, every six hours, daily, or weekly work. App event watches one exact connected source. Webhook creates a private URL plus a one-time HMAC secret.
- GitHub: new issues or failed workflow runs in one consented repository.
- Slack: a member command or matching message in one consented channel.
- Gmail and Outlook: messages returned by one bounded mailbox query.
- Microsoft Teams: a member command or matching message in one selected Team channel.
- Sentry: issues returned by one bounded organization query.
Set guardrails
Every hosted trigger has a daily run cap and replay window. Signed webhooks can also match one safe JSON field before spending a run. Optional quiet hours use the selected IANA timezone; matching events are encrypted, held, and resumed after quiet hours rather than discarded.
- Duplicate delivery IDs do not create a second run.
- At most two runs for one automation may be active at once.
- Provider checks do not call a model until a new matching event is claimed.
- Bot-authored Slack messages are ignored to prevent notification loops.
- Pausing an automation stops schedules, provider checks, tests, and inbound webhooks.
Test and inspect in Projects
Expand an automation in Projects to send a manual trigger test through the same policy, review the latest 50 trigger deliveries, read the latest run log, pause or resume work, and inspect trigger health. If that automation has a signed result webhook, the same view can send a signed result test, replace the endpoint, rotate its one-time secret, disable delivery, and retry a failed terminal event without rerunning the Agent Team. Trigger receipts and result status are labeled separately. Delivery receipts are retained for 30 days. Green is healthy, amber is delayed, and red needs attention.
Watch one global inbox
The bell in the global toolbar shows running, completed, failed, approval-needed, and Team Health work. Open Agent Team Inbox for the complete queue, private receipts, source-delivery state, and recoverable reply failures. Green means active or newly completed work, amber means an approval or health issue needs review, and red means a task failed or Team Health found a critical condition. Clearing a finished task permanently dismisses that task for the account; durable run receipts remain in the Inbox.
Deliver results
Configure email, Telegram, a per-automation result webhook, or Reply in the source conversation. Slack and Microsoft Teams post inside the original thread; Gmail and Outlook use their native reply APIs on the exact source message. Every reply contains only status, approximate cost, and an authenticated private receipt link. Background-task email turns bounded structured results into a short human-readable summary and returns to the exact Agent Team, board, or architecture project where the task started. Webhook deliveries use a stable delivery ID and an HMAC signature so receivers can verify and deduplicate retries. Local Lite, open-tab schedules, and ordinary browser task results never use a hosted result destination.