Design Partner Program · Now onboarding

Webhooks that arrive.
In order. Exactly once.

HookSift is the delivery layer between your event bus and your customers' endpoints — ordered per subscription, deduplicated at the endpoint, with per-endpoint retries you never have to write.

Apply for Access Read the docs →

payments · scheduling · CRM platforms

Delivery Ledger Live

event · orders.paid evt_9Fh2kQ

✓ acme.co/webhooks try 1 DELIVERED

✓ api.storely.io/wh try 1 DELIVERED

↻ hooks.dev-app.com/ep try 2 RETRYING

✕ old.legacy-host.net/x try 5 DEAD-LETTER

2 delivered·1 retrying·1 quarantined·Ordered ✓

Illustrative · one event → four endpoints

Ordered·Exactly-once·Signed·Replayable·Per-endpoint retries·Tenant-isolated·Ordered·Exactly-once·Signed·Replayable·Per-endpoint retries·Tenant-isolated·

4,182,904

Events dispatched this week live

Trusted by payments, scheduling, and CRM platforms to deliver events their customers actually receive — no duplicates, no reordering, no silent drops.

The last mile

Your event bus is fine. The last mile is where webhooks die.

The internal bus — Kafka, an outbox, SNS — usually works. The failure is fanning one event out to N customer endpoints that are slow, flaky, and outside your control. Most teams solve it with a for loop and rediscover the same three bugs in production.

✕ Duplicates

Two app servers process the same event, or a retry fires after a timeout that actually succeeded. The webhook lands twice — and so does the charge, the email, the provisioning call.

✕ Out of order

subscription.updated arrives before subscription.created. The receiver builds state from events in the wrong sequence, and their data is quietly wrong.

✕ Silent drops

An endpoint returns a 500, the sending loop moves on, and the event is gone. No delivery log, no replay — just a support ticket you can't prove either way.

HookSift makes all three impossible by construction. ↓

What you get

Delivery guarantees, not delivery attempts.

01 Per-key order

Ordered per subscription

Events land in publish order for each endpoint. No reordering, no reconciling out-of-sequence state on the receiving side.

02 Idempotent

Exactly-once at the endpoint

Every delivery carries an idempotency key. HookSift dedupes redeliveries so a retry never fires a duplicate side effect.

03 Backoff + DLQ

Retry policies per endpoint

Set max attempts, backoff curve, and timeout per subscription. Dead endpoints are quarantined to a dead-letter, not retried into the ground.

04 No HOL blocking

Fair fan-out

One tenant's slow endpoint can't stall delivery for anyone else. Per-tenant isolation prevents head-of-line blocking by default.

05 HMAC + TS

Signed, verifiable payloads

Every request is HMAC-signed with a per-endpoint secret and a timestamp. Receivers verify authenticity and reject replays.

06 Log + replay

Replay without republishing

Every attempt is logged with status, latency, and response body. Replay one event or a time range straight from the delivery log.

How it works

Publish once. We handle the fan-out.

Publish

POST an event to your tenant's dispatch channel, or point HookSift at your existing bus. One call, one event.

Fan out

HookSift resolves every subscribed endpoint, signs the payload, and delivers in order — retrying failures on each endpoint's own policy.

Guarantee

Each endpoint sees the event exactly once, in sequence. Every attempt is logged and replayable. You write zero delivery code.

dispatch.sh

# Publish an event — fanned out to every subscribed endpoint
curl -X POST https://events.hooksift.com/dispatch/tenant/m8r2k4/channel/orders \
  -H "Authorization: Bearer $HOOKSIFT_KEY" \
  -H "Hooksift-Idempotency-Key: evt_9Fh2kQ" \
  -H "Content-Type: application/json" \
  -d '{ "type": "orders.paid", "data": { "order_id": "o_5521", "amount": 4900 } }'

# 202 Accepted
# { "dispatch_id": "dsp_7yQ2fD", "endpoints": 4, "status": "queued" }

Root GET / returns 403 Forbidden from hooksift-edge — the Dispatch API only answers signed, authenticated requests. See the docs.

Why not build it

You could build this. You'd build it six times.

Every service that emits events reinvents the same delivery layer. HookSift is that layer, once, with the guarantees written down.

Capability	HookSift	Roll-your-own	General MQ
Ordered per endpoint	✓ Guaranteed	Best-effort	Coarse (partition)
Exactly-once at endpoint	✓ Sender dedup	On the receiver	On the receiver
Per-endpoint retry / DLQ	✓ Configurable	Hand-rolled	You build the worker
Tenant isolation	✓ Default	—	Manual per-partition
Signed payloads	✓ HMAC + TS	DIY	DIY
Replay + delivery log	✓ Built in	Grep logs	—

Pricing

Priced per delivery. Onboarded by application.

No self-serve signup during the design-partner period. Every plan is provisioned by our integrations team.

Signal

By application

Free during design-partner period

→ 1 tenant
→ Up to 100k deliveries / mo
→ 3 endpoints
→ Standard retry policies
→ 7-day delivery log
→ HMAC signing

Apply for Access

Fan-out

★ Popular

Most partners start here

$400/mo

Indicative — set per cohort

→ Up to 5M deliveries / mo
→ Unlimited endpoints
→ Custom retry curves + DLQ
→ 30-day log + replay
→ Per-tenant isolation SLO
→ Signature rotation

Apply for Access

Dispatch

Contract

Custom

Volume & multi-region

→ Multi-region edge delivery
→ Dedicated fan-out capacity
→ 90-day log
→ Priority incident channel
→ Delivery guarantees in contract
→ SSO + audit log

Talk to Integrations

Design Partner Program

Apply for access

We onboard new platforms in small cohorts. Applications are reviewed bi-weekly by our integrations team.

Work email

Company / platform

Monthly event volume

Your event bus

Delivery use case

No credit card. No demo call required. We read every application.

Stop writing delivery code.

Apply to the HookSift design-partner cohort and hand the last mile to us.