Billing and balance
One prepaid balance for every project in a team. Requests draw from it as they run.
Prepaid balance
| Amount | Meaning |
|---|---|
| Posted | Credit added, minus settled charges. |
| Reserved | Held for requests that are still running or waiting for their final usage. |
| Available | Posted minus reserved. New requests must fit in this. |
Read all three with GET /v1/balance. The team balance also shows at the bottom of the console sidebar.
Reservations
Before a request is sent to a provider, its worst-case cost is reserved from the available balance, based on the prompt size and the output token limit. The hold covers every provider attempt the request may make (see Routing), and every web search it may run. When usage is known, usually within seconds, the actual charge is posted and the rest is released. Setting max_tokens keeps reservations small. Without it, the output limit is 4,096 tokens, or your maxOutputTokens Directive, and that much is reserved.
If available balance can't cover the reservation, the request is refused with 402 insufficient_balance before any provider is called. Many requests in flight at once each hold their own reservation, so keep enough available for all of them.
Adding credit
Owners, admins and billing admins add credit in the console under Billing, from $25 to $5,000 at a time. Payment is through Stripe Checkout; the balance updates once Stripe confirms the payment. GET /v1/billing/topups lists recent top-ups.
Starter credit
A new team that hasn't added credit can claim a one-time $5 starter credit, from its project's Get started checklist or from Billing, to try its first requests before paying. An owner, admin or billing admin with a verified email saves a card in Stripe Checkout to claim it: Stripe checks the card, nothing is charged, and the card isn't kept. Each team and each card can claim it once, and prepaid cards can't.
The starter credit lasts 30 days. Spending counts against it first, and whatever of it is unspent then leaves the balance; credit you added yourself is never taken.
Invoices and billing details
Every top-up comes with a Stripe invoice and a card receipt, under Invoices and top-ups on Billing. Checkout asks for the billing address and, optionally, a tax ID such as a VAT number or EIN, and keeps them for the next time. Billing details shows the company name, address, tax IDs and the email invoices go to; Edit in Stripe changes them, and a PO number set there is printed on every invoice from then on. Top-ups made before invoices existed have a receipt only.
Once auto reload has a saved card, Pay now under Add credit charges it in one click. If your bank asks to confirm the payment, you're sent to it and back. Pay with another method still opens Checkout.
Spend and export
Billing → Spend and export shows what the team spent over any range of days, grouped by project, model, key or day, with the range's top-ups and their invoices. Download CSV has one row per charge to the balance: each settled request with its project, key, model and tokens, and each add-on fee, such as the Auditor white label.
Monthly statements and invoicing
On the 1st of each month, a team that paid monthly add-on fees from its balance, such as the Auditor white label, gets one invoice listing last month's fees, already marked paid. It's under Invoices and top-ups with the same billing details and PO number.
Teams that need to pay by invoice can ask us for postpaid billing. With it, requests keep running past zero up to an agreed credit limit, and on the 1st the team gets an invoice for what it owes, due in 30 days, payable by card or bank transfer from the invoice's page. Paying it brings the balance back up. If an invoice is more than 7 days overdue, new requests are refused with 402 insufficient_balance until it's paid, and billing contacts are emailed.
Auto reload
Turn on auto reload to buy a set amount of credit whenever available balance drops below a threshold, using a saved card. The defaults are a $100 reload when the balance falls below $20. A monthly limit (and, if you set one, a daily limit) caps how much it can buy; each must be at least the reload amount. If a charge fails, auto reload pauses and the team's owners and admins are emailed.
Budgets and alerts
A balance stops spending only when it runs out. To cap spend earlier, set monthly or daily budgets on the team, a project, an environment or a key in Directives; a request over budget is refused with 402 budget_exceeded. Alerts and webhooks tell you when a budget is nearly spent or the balance runs low.
What you pay
Each request costs what the provider charged for it, at its list price for the tokens used (and any web searches), plus a platform fee of 7.5% on that cost. There are no per-model markups and no fee for requests that are refused before dispatch. The receipt for every request shows the upstream cost, the fee and the charge. Current prices are on Pricing and from GET /v1/pricing.