Skip to main content
Outbound webhooks push platform events to your own HTTP endpoints — the reverse direction of inbound webhooks. Use them to get notified when deploys finish, machines restart, backups fail, or an agent run needs your attention. Outbound webhooks are managed via the API (there is no CLI command yet):

Events

The agent events make outbound webhooks the machine-readable counterpart of the Inbox: agent_run.needs_attention fires when a run pauses on an escalation, and agent_message.created fires on every send_message.

Verifying deliveries

Each delivery is signed with the webhook’s secret (returned once, at creation): Verify by computing the HMAC of the raw request body with your secret and comparing in constant time.

Delivery semantics

  • Non-2xx responses (and transport errors) are retried up to 8 attempts with growing backoff (30s → 2m → 10m → 1h → 6h).
  • Delivery history is retained for 14 days and can be inspected — and individual deliveries re-sent — via the API.
  • Endpoint URLs must be public http(s) on ports 80/443; private and internal addresses are rejected.