Questions, answered.

Paid plans meter accepted events beyond the included allowance at the listed overage rate. A safety ceiling can reject new requests to limit runaway usage. Already accepted events remain subject to their retention window.

On the free plan we warn you at 90 percent and keep receiving and delivering up to double the included volume. Past that, new events get a 429 at the edge and provider retry behavior varies. Check the provider’s redelivery tools before assuming those events will return. Nothing already accepted is affected.

A successful acceptance response follows durable event storage and pending dispatch work. Delivery is at least once, and your application should handle duplicate requests.

Events leave storage only when they age out of your plan's retention window, or when you delete them.

Every event is stored the moment it arrives, with its raw headers, raw body, and every delivery attempt. The free plan keeps 15 days, Starter 30, Pro 60, and Business 180.

Anything inside that window can be replayed, one event at a time or a whole filtered range. That is the point of the window: when you fix the bug that broke your handler, the events that hit it are still there to run again.

Automatic retries continue until the configured attempt limit or event retention is reached. Paid plans can configure up to fifty attempts with up to a day between attempts.

After ten terminal failures at a destination, a circuit breaker holds new attempts until a probe is allowed. A successful probe clears that destination’s circuit. Retention still bounds how long events can wait.

You hear about it before that happens: failure alerts fire by email after five consecutive failures by default.

Delivery is asynchronous. Queue load and destination response time affect latency. Retries follow your backoff setting, so a failing destination sees increasing gaps rather than a tight loop.

On Amazon Web Services. Event records sit in DynamoDB, oversized payloads in S3, and delivery runs through SQS and Lambda. Nothing is copied to a third party for analytics.

Yes. Request headers, body, signature result, and delivery attempts are available during retention. Configured privacy fields are masked in inspection, and response excerpts are hidden when masking is active.

Yes, on Pro and Business. Paste a Slack or Discord incoming webhook URL in alert settings and failure alerts arrive in that channel in its own format, with a link to the endpoint. Any other HTTPS webhook works too. Email alerts are on every plan, and a test button checks each channel before you need it.

Yes. The shellorbit-listen script forwards each new event an endpoint receives to a URL on your machine, with the original headers and body, so provider signatures still verify. It polls the API with your key, so there is no tunnel to run. Your destinations keep receiving events as usual.

Point the provider back at your own URL and delivery stops being ours. Export your stored events over the API before you cancel if you want to keep them.

Yes. First time subscribers get a 7 day free trial on Starter, Pro, or Business. A payment method is needed to start it, nothing is charged during the trial, and cancelling before it ends costs nothing. We email you about two days before it ends. The free Hobby plan stays free with no trial needed.

From the dashboard, at any time. The plan runs to the end of the paid period and then stops. Payments already made are not refunded, which is set out in the refund policy.

Every plan, including the free one, carries a 99.9 percent monthly commitment on the ingest endpoint, written into the terms. Paid plans are credited for a shortfall; the free plan has no fee to credit, so there the commitment is a published target.