One job, one credit.

A credit is consumed when a job is created, and recorded against that job, so the ledger is auditable per job rather than as a running counter you have to trust.

A job that fails outright is refunded automatically and marked as refunded. You are not billed for our errors, and you do not have to notice them in order to avoid being billed for them.

Reads are free. Polling a job, listing jobs and checking your balance cost neither a credit nor quota, so poll rather than holding connections open.

Repeat requests for the same subject inside the cache window do not create a second billable job. Nor does a request that arrives while an identical one is still running: it joins the job already in flight rather than starting a second. When you do want a genuinely fresh run, ask for one explicitly and it bills.

Monitoring re-checks on an existing campaign are not separately billed. The re-check loop is part of the campaign, not a new one, so the recurring product you sell does not cost you more each month it keeps working.

Answering a broker that has asked for more information costs nothing either. Charging for the step that rescues a stalled removal would be an odd way to price a removal product.

Every key has a reconciliation ledger showing jobs, credits consumed, refunds and the reason for each. It is exposed through the API, not assembled by us on request.

Rates are quoted per partner, because the things that move them are specific to how you would use the API: monthly volume, whether you commit annually or run month to month, and whether you need scan only or scan and removal.

Tell us roughly what those look like and you will get a straight number back, not a discovery call.