Credits and Resource Usage

There are only two words to learn here, and they are not two counters.

• Credits are what you hold. A plan grants them; you can buy more at any time.
• Resource Usage (RU) is what things cost. Every unit of it comes off your credit balance.

So there is one balance, drawn down by everything you do. Nothing is billed separately, nothing is metered on a second axis, and there is no per-environment charge sitting underneath.

Where it actually goes

This is the part worth internalising before you model anything, because the intuition from other platforms is wrong here.

	What it costs
A call through the gateway	A fraction of one unit
Designing, publishing, comparing a specification, compliance checks, impact analysis	Whole units, and the reason your balance moves

Serving requests is the cheap part. That is not a discount, it is what the platform is: the expensive work is the thinking — understanding a contract, deciding whether a change breaks somebody, producing the evidence — and that happens a handful of times per release rather than once per request.

The practical consequence is that traffic growth and cost growth are not the same curve. Ten times the consumers on the same API does not cost ten times as much to run, because the API was designed, compared and published once. If you are used to per-call gateway pricing with an environment fee on top, this is the difference worth checking against your own numbers.

Why a stand-in is cheap to hammer

A mock is answered by the same path as anything else, so a partner integrating against dynamic mock data for a fortnight costs a fraction of a unit per call and nothing else. Evaluations do not need rationing.

The same unit, in both directions

RU is a unit of API consumption, and it measures consumption wherever it happens:

• Your consumption of Apiway — drawn from the credits you hold, as above.
• Your consumers’ consumption of you — the meter you bill from. You set the RU cost per operation, and an OpenSLA tier grants each consumer a quota. See Metering & Billing.

These are separate ledgers that happen to share a unit, which is deliberate: the thing being counted is the same thing. Your customer’s quota has nothing to do with your balance, and exhausting one does not touch the other.

Running out

Like anything you pay for, when it is used up the work stops — but you should never be surprised by it, and your traffic does not have to stop with it.

You see it coming. Risk management watches your consumption against your tier’s allowance. If you keep reaching the upper limit, it tells you — well before anything runs out — so you can plan the move to the next tier on your own schedule rather than discover it in production.

A reserve takes over. If you keep a pay-as-you-go reserve, Apiway draws on it when your tier’s credits are spent, and everything carries on. Nothing pauses, nothing is renegotiated in a hurry.

Only then does work stop. With no credits and no reserve, the platform stops doing chargeable work rather than doing it and invoicing you afterwards. Top up and it resumes. The balance is yours and is not reset on you — what you buy stays until it is spent.

It is the same arrangement you offer your own consumers, on the same meter: an allowance per tier, and a bigger tier when they outgrow it. It is how you get paid, and it is how we get paid — we run Apiway on it ourselves.

A consumer exhausting their quota is a different event, and they receive 402 at the gateway. See Metering & Billing and Cost Control.

See also: Metering & Billing · Cost Control · OpenSLA