Every webhook platform promises delivery. The Svix Message Attempt API is where Svix proves it. Its description is one line, “Attempts to deliver Messages to Endpoints”, and its nine operations are the audit trail behind that line.
Of the 29 API pages Svix has on the network, this is the one I would read first, because it is the part of a webhook product most companies never expose.
What is in it
| Method | Operation |
|---|---|
| GET | List Attempts By Endpoint |
| GET | Count Attempts By Endpoint |
| GET | List Attempts By Msg |
| GET | List Attempted Messages |
| GET | List Attempted Destinations |
| GET | Get Attempt |
| GET | Get Attempt Headers |
| DELETE | Delete attempt response body |
| POST | Resend Webhook |
Seven reads, one delete, one resend. You can walk the delivery graph from either side, which messages an endpoint was sent, which endpoints a message went to, and down to the headers of a single attempt. Then you can resend one message to one endpoint.
Three things worth noticing
The delete is a privacy control, not a cleanup. DELETE .../attempt/{attempt_id}/content removes the response body a customer’s endpoint sent back, and leaves the attempt record intact. Receivers sometimes echo sensitive data into webhook responses. Svix gives you a way to expunge it without destroying the evidence that delivery happened. That is a small operation carrying a large design decision.
Resend is scoped as narrowly as it can be. POST /api/v1/app/{app_id}/msg/{msg_id}/endpoint/{endpoint_id}/resend takes a message and a destination. It does not replay a queue or retry an endpoint wholesale. For an agent, that is exactly the right unit. The one write that causes a side effect is bounded to one delivery.
The contract knows where the data lives. The OpenAPI (3.2.0, version 1.922.0) declares five regional servers: EU, US, Canada, Australia and India. Attempt data is delivery data, and delivery data about your customers has a jurisdiction. The page’s ErrorCodes property points at Svix’s retry documentation, which is where attempt statuses get their meaning.
Where it sits in the score
Svix scores 76.7, exemplar, with operational transparency at 81.1 and contract governance at 45.5. Agent Readiness is 49.4, agent-native, with reversibility, idempotency and error semantics all verified, which is the combination this API depends on. Every one of Svix’s 28 contracts is callable and none is derived.
Two of the 15 Arazzo workflows on the Svix provider page, “Resend a Failed Message Attempt” and “Send Message and Confirm Delivery”, are built around this API, which is a fair signal of where the real integration work happens. One caution: the agentic access profile (230 operations, 134 acting, 1 human-in-the-loop) is derived by us, not published by Svix, and earns it no credit here. One human-in-the-loop marker across 134 acting operations is thin, and the resend would be a sensible place to add a second.
Takeaway
If you are evaluating any webhook provider, ask for this API first. A platform that lets you list, count, inspect, expunge and resend individual delivery attempts is telling you it expects to be audited. Svix does.
Read the reference at api.svix.com/docs, and the API page at apis.io/apis/svix/svix-message-attempt-api/.