See webhook deliveries and retry failures

Every webhook attempt is logged per endpoint, so you can see what was sent, whether it succeeded, and why it failed. Retries are automatic — this page explains the log, the delivery states, and how to recover a failing endpoint.

Open the deliveries log

As the account owner, open Settings → Integrations → Webhooks.
On the endpoint you want to inspect, click Deliveries. A table opens showing the most recent attempts.
The columns are Event, State, Code (the last HTTP status your endpoint returned), Tries (attempts so far), and When.
Click Hide log to collapse it. No deliveries yet. means nothing has been sent to that endpoint.

Each endpoint row also shows its last delivery at a glance — for example last: delivered (200) or last: dead.

What the states mean

pending — queued or mid-retry; not yet acknowledged with a 2xx.
delivered — your endpoint returned a 2xx.
dead — all retries were exhausted without success; this delivery will not be attempted again.

How automatic retries work

The first attempt is immediate. A failed delivery — a non-2xx status, a network error, or a response slower than about 10 seconds — is retried up to five more times with backoff:

after ~30 seconds
after ~2 minutes
after ~10 minutes
after ~1 hour
after ~4 hours

That is six attempts over roughly five and a quarter hours; after the last one fails, the delivery is marked dead. There is no button to replay a single past delivery: fix your endpoint so future events flow, or use Send test to push a fresh signed event on demand.

When an endpoint keeps failing

An endpoint that fails repeatedly can be turned off automatically — its row then shows a disabled (failures) chip, and while disabled it receives nothing. After you fix your receiver:

Click Send test to confirm it now returns a 2xx (the row shows the live status code).
Click Enable to turn the endpoint back on. New events resume immediately.
If you changed infrastructure, double-check the Endpoint URL is still correct and roll the signing secret if it may have leaked.

Removing an endpoint

Remove deletes the endpoint. Any still-pending deliveries queued for it are marked dead, because they can never succeed once the endpoint is gone.