Outbound webhook delivery
“We should offer webhooks.”
It sounds like one endpoint. It is this list — and the half of it nobody remembers until a customer asks why an event from last Tuesday never arrived.
- HMAC signing, with a scheme receivers already know
- A timestamp and delivery id inside the signature, or it replays
- Exponential retries that survive a receiver being down all day
- Jitter, or every queued delivery stampedes them on recovery
- Knowing which failures are worth retrying and which are not
- Secret rotation with an overlap, because nobody can switch in one instant
- An append-only log of every attempt and response
- A way for your customers to see that log without emailing you
HookPulse is all of it, behind one POST.
No card. 2,000 deliveries on the trial.
One request in. We own everything after it.
Your handler returns immediately — a webhook should never put somebody else's slow server inside your own request. From there it is our problem: signed, delivered, retried, and written down.
See the verification snippetawait fetch('https://hookpulse.altixcode.com/api/v1/apps/APP_ID/events', {
method: 'POST',
headers: { Authorization: 'Bearer hp_…' },
body: JSON.stringify({
type: 'invoice.paid',
payload: { id: 'in_123', amount: 4900 },
dedupeKey: 'invoice-123-paid',
}),
})Why not build it yourself?
Most teams start to. The signing and the first retry are a day's work. The rest — jitter, replay windows, rotation overlap, a customer-facing log — is the part that gets cut, and it is the part your customers actually notice.
The whole list, not the first half
Signing and one retry get built in an afternoon by almost every team that tries. Jitter, retryability rules, rotation overlap and a customer-readable log are the part that gets cut under deadline — and the part a customer notices when it's missing.
Your endpoint URL is re-checked before every attempt, not just saved
A hostname that resolves into your own private network is refused at save time and refused again, by resolved address, immediately before each delivery attempt — and a redirect is never followed to a host you didn't name. A webhook sender that only checks once is a way for a customer's misconfigured or malicious endpoint to probe your infrastructure.
A log your customers can see, not just one you can
The delivery log ships as a scoped, signed, time-limited API today, built to be embedded in your own settings page — so "why didn't my webhook arrive" stops being a support ticket you have to answer by hand.
The retry schedule, stated rather than implied.
Six attempts, stretching over more than seven hours, each jittered so a receiver coming back up is not hit by everything at once. Then we stop and tell you — a delivery that has failed all day needs a person, not another request.
- immediately
- +15sattempt 2
- +2mattempt 3
- +15mattempt 4
- +1hattempt 5
- +6hattempt 6
Priced by delivery.
An event fanned out to five endpoints is five deliveries, because that is five requests we made and five sets of retries we are responsible for. Retries of the same delivery are not counted again.
Scale
A marketplace, or an API with real traffic.
$149/ month
1,000,000 deliveries / month
Start with Scale