Webhook & Event Subscription Management maps to 658 providers and 818 API surfaces, with 350 providers rated strong or better. Registering a webhook is close to a commodity. Recovering one that failed is not. A scan of those providers’ operations finds only 85 with a write operation on a path that resends, retries, replays or redelivers something event-shaped.
The scan by band
| Band | Providers | With a recovery operation | Share |
|---|---|---|---|
| Exemplar | 137 | 18 | 13% |
| Strong | 213 | 27 | 13% |
| Developing | 176 | 29 | 16% |
| Thin | 108 | 7 | 6% |
| Emerging | 23 | 4 | 17% |
| Minimal | 1 | 0 | 0% |
Treat 85 as a ceiling on what the contracts show. It matches on path words, and a few hits are not webhook recovery, such as Cacheflow resending an email and Hover resending notifications for a capture request. It also can’t see a provider that retries automatically, or that offers replay only in a dashboard. What it measures is whether a caller can trigger recovery through the contract.
The Kin Score does not separate this. Exemplars do it at 13%, developing providers at 16%, and most of every band has no such operation on the record. A high score does not tell you a provider has thought about the failed event.
What good looks like
The strongest shapes treat delivery as a resource. Svix has PUT /api/v1/event-type/{event_type_name}/retry-schedule, plus bulk-replay and replay-missing on an endpoint. Convoy offers batchretry and forceresend on event deliveries. Fireblocks has POST /webhooks/resend, FirstPromoter and Drippay retry a specific delivery by ID, and GitLab and PayPal resend a single event. Only two providers have a dead-letter path naming webhooks: Kita with GET /api/v1/webhooks/dlq and CertifyOS with POST /webhooks/dlq/drain.
That matters more for agents than for people. An agent that subscribes to events and misses one has no human looking at a dashboard. A contract operation is the only recovery path it can find by itself.
What is ours to disclose
The model breaks this capability into three children: Webhook Subscription Lifecycle, Event Schema Catalogue, and Webhook Delivery Reliability Management. All three show no catalog coverage on the page. The mapping stops at level two, so our catalog says nothing structured about delivery reliability, which is the question that matters most. The scan above is a stand-in until it does.
The agent scores at the top also need a disclosure. Xquik leads at 86.9. Next come Yarnhen and Yawplet at 78.8, then Eventwren and Hagglebee at 73.3. All four are agent-first publishing sites operated by API Evangelist, which is us, and they sit in the exemplar band of a ranking we publish. Read them as our own reference implementation and not as independent market evidence. Telnyx (87.6, 72.7) and OpenAI (82.6, 58.0) are the stronger independent signals.
Takeaway
When you evaluate a webhook provider, skip the question of whether webhooks exist, since 658 providers say yes. Ask what happens to the event that failed at 3am, and whether you can fix that with an API call. If you publish webhooks, a list-deliveries and retry pair is a small addition that few of your peers have.
See the capability at apis.io/capabilities/webhook-event-subscription-management/.