What happens if one target is down but the others are up?

The healthy targets are delivered as normal. The failing one is retried on its own schedule and, if it still cannot be reached, parked in your dead-letter queue to inspect and replay — one slow or broken target never holds up the rest.

Are the deliveries parallel or sequential?

Parallel. Every target receives the request at the same time, each with its own authentication, headers, and retry budget.

Do I get one response back, or many?

Fan-out is asynchronous by default — you get an accepted acknowledgement and delivery happens in the background. If you need a synchronous answer, mixed-mode returns one nominated target's response in the same call.

How are fan-out deliveries counted for billing?

One delivery per target, so a request to three targets is three deliveries. Deliveries that fail on our side are never charged. The pricing page has the full model.

Fan-out

One in.
Everywhere out.

One request fans out to every service you connect. We authenticate, validate, bill, and deliver — you ship.

What you get

Delivery you can trust.

At-least-once delivery.

Every target gets the request. Failed deliveries are retried, then held in a queue you can replay.

Retries, handled.

Transient failures retried with exponential backoff. No thundering herd.

A failure queue you can replay.

What can’t be delivered waits in a queue you can inspect and replay. You decide when to retry.

Reshape in the middle.

Direct, mapped, or transformed per target. Each service gets exactly the shape it expects.

Many out, one back.

Mixed-mode: fan out to every target and return one target’s response in the same call.

How it works

One call, three steps.

You send one request

Your app POSTs a single request to your EchoRelay endpoint with your project key.
We authenticate and fan it out

We verify your key, validate the payload, meter it, then deliver to every target you connected — in parallel, each with its own auth, headers, and retry budget.
Every target is delivered

Each target gets the request at-least-once, retried on failure. Anything still undeliverable waits in a queue you can inspect and replay.

Each delivery is 1 credit — a request to 3 targets is 3 credits, plus 1 per 64 KB. Deliveries that fail on our side aren’t charged.

Questions

Fan-out, answered.

What happens if one target is down but the others are up?: The healthy targets are delivered as normal. The failing one is retried on its own schedule and, if it still cannot be reached, parked in your dead-letter queue to inspect and replay — one slow or broken target never holds up the rest.
Are the deliveries parallel or sequential?: Parallel. Every target receives the request at the same time, each with its own authentication, headers, and retry budget.
Do I get one response back, or many?: Fan-out is asynchronous by default — you get an accepted acknowledgement and delivery happens in the background. If you need a synchronous answer, mixed-mode returns one nominated target's response in the same call.
How are fan-out deliveries counted for billing?: One delivery per target, so a request to three targets is three deliveries. Deliveries that fail on our side are never charged. The pricing page has the full model.

Not the right shape? If you call a single target and want its response streamed back token by token, use Streaming. To keep a shadow copy of every request for migration or audit, use Mirror.

Connect everything. Build nothing.

Start free Built for AI

No credit card required.