Sign up
Docs menu

Billing

Sator is prepaid. You add credit in US dollars, every request debits your balance at the per-token rates on the price page, and nothing recurs.

Credits#

  • Minimum top-up: $10. Paid by card through Stripe from the dashboard.
  • Every top-up has a receipt. The Receipt link under Receipts on the dashboard opens Stripe's page for the purchase, where the receipt and the invoice are both downloadable as PDF, with the billing address you gave at checkout.
  • Credits never expire. They are non-transferable and redeemable only against Sator usage.
  • Price changes apply to future usage only. Your balance is a dollar balance and is never reduced by a price change.
  • There is no auto-top-up and no saved card: the Add-credits button is the only way a balance goes up, which caps what a leaked key can spend at the balance behind it.

Refunds and payment reversals are covered in the Terms of Service.

What a request costs#

Each response's usage block is what you are billed for: input tokens at the input rate, output tokens at the output rate, and cached input tokens at the cache-read rate instead of the input rate. Where a model has a cache-write rate, tokens written to its cache are billed at that rate; where it has none, they are not billed. Rates are per million tokens; the debit is tokens × rate, rounded to a millionth of a dollar.

Four models — qwen3.7-plus, qwen3.6-plus, grok-4.6, gpt-5.6-luna — have a second rate set that applies to the whole request, output included, when the input side (input plus cache reads plus cache writes) exceeds 200,000 tokens. Both rate sets are on the price page.

Every debit is recorded under the request's x-request-id. The dashboard shows spend by day, by model and by key, from the same records.

Zero-balance cutoff#

Service stops immediately when your balance reaches $0, including part-way through a response. A request that arrives at a balance of $0 or below is refused 402 insufficient_credits before it reaches a model. A streamed response that exhausts the balance mid-stream stops there. Add credit and retry.

Because a response is billed after it is generated, a single large request can take the balance slightly below zero; that overshoot is bounded by the cost of one maximal request on the model used.

Interrupted responses#

If you disconnect mid-stream, the generation still completes and is billed. If a stream is cut short for any other reason, the tokens up to the cut are billed from an estimate, flagged as estimated, and corrected in a nightly reconciliation against the model's own count.

Low-balance email#

When a debit takes your balance below $2, you receive an email saying so, to the address on the account — at most once every seven days, whatever happens to the balance in between, so a small top-up spent back through $2 the same afternoon does not mail you again.

Keys cost nothing#

Creating a key needs no balance. Only calls cost money, so you can set up every tool at $0 and pay when you are ready.