Delivery that does not give up
Eight attempts over thirteen hours, exponential backoff with jitter, per-endpoint rate limits so one slow consumer never backs up the queue, and circuit breaking on endpoints that are clearly dead.
Yaranex is delivery infrastructure for API companies. One API call fans out to every customer endpoint, with retries, signing, rate limits, and a searchable log of every attempt. When a customer asks "did you send it," the answer takes ten seconds.
One endpoint to integrate: POST /v1/events. Nothing to deploy.
Typical engineering time to ship a first in-house delivery system. Retries, key rotation, and a usable log come later, if ever.
Integration support tickets that trace back to a failed or missing webhook, based on our interviews with platform teams.
First delivery attempts that fail on a normal day. Customer endpoints go down. Handling that is the whole job.
A retry loop in a worker queue is not delivery infrastructure. Yaranex is the version you would build with two years and nothing else on the roadmap.
Eight attempts over thirteen hours, exponential backoff with jitter, per-endpoint rate limits so one slow consumer never backs up the queue, and circuit breaking on endpoints that are clearly dead.
Every outbound event carries an HMAC signature, with zero-downtime key rotation. Inbound, we verify third-party webhooks, dedupe by event id, and hold them for replay so a bad deploy never eats a payment notification.
Full payload search across every event. Every attempt is recorded with request, response, and timing. And your customers get their own endpoint dashboard, so they debug their side without opening a ticket on yours.
Also in the box: payload transformations per endpoint, so customers on your v1 schema keep working while you ship v2. No versioning migration emails, no broken integrations.
You emit events. We get them where they need to go and keep the receipts.
One call per event. We fan out to every endpoint the customer has subscribed, apply their transformations, and sign each request.
Retries, backoff, rate limits, circuit breakers, key rotation. An endpoint being down at 3am is a line in a log, not a page to your on-call.
Each customer gets a dashboard for their own endpoints: attempts, response codes, latency, and a replay button. The "did you send it" ticket stops existing.
# one event in, delivered everywhere it belongs curl -X POST https://api.yaranex.com/v1/events \ -H "Authorization: Bearer $YARANEX_KEY" \ -H "Content-Type: application/json" \ -d '{ "customer": "cus_8fk2m1", "type": "invoice.paid", "payload": { "invoice_id": "inv_20260142", "amount_cents": 490000, "currency": "usd" } }'
Per delivery, not per seat. Start free, pay when your volume says you should.
Early access pricing is locked for 12 months for design partners. Overages billed per delivery, never surprise-tiered.
We guarantee at-least-once delivery and best-effort ordering per endpoint. Under retries, strict global ordering is a fiction, and vendors who promise it are hiding a queue that stalls behind one bad endpoint. Instead, every event carries a monotonic per-subject sequence number in its headers, so consumers that care about order can detect and reorder cheaply. Most never need to.
Retention is 3 days on Build, 30 on Scale, 90 or custom on Enterprise. Payloads are encrypted at rest with per-tenant keys and in transit with TLS 1.3. There is a deletion API for compliance requests, and you can mark fields for redaction so secrets never land in the log at all.
We retry 8 times over roughly 13 hours with exponential backoff and jitter. After repeated failures, the endpoint is marked degraded, its owner sees that on their dashboard, and you can both replay the missed window with one action once it recovers. Sustained failures trip a circuit breaker so we stop hammering a dead host.
Both. Ingest gives you a stable URL for each third-party source. We verify signatures, dedupe by event id, and buffer everything for replay, so a bad deploy on your side never loses a payment notification. Then we deliver into your service with the same retry engine.
Because you built delivery, not observability, and the tickets prove it. Migration is boring on purpose: keep your event names and payloads, point your emitter at POST /v1/events, and dual-write during cutover. Your customers change nothing. Endpoint secrets can be imported so existing signature verification keeps working.
We are onboarding a small group of design partners now. If webhooks are load-bearing for your product, we want to talk this week.
Request early access