For managed requests, SingleAPI uses its provider key and charges your SingleAPI wallet. On Overview, check your credit balance. If Add credits is available, enter an amount and complete the hosted checkout. Credit is added after payment is confirmed; returning from checkout alone does not confirm it. BYOK requests use your provider account and have zero SingleAPI charge.
SingleAPI may temporarily hold enough credit for a managed request. The final charge is based on valid provider usage and cannot exceed that hold; unused credit is released. If complete provider usage is unavailable, SingleAPI does not invent a token count or charge from an estimate. Your dashboard balance may show rounded cents, while request detail has a more precise amount.
Pricing and units
The signed-in model catalogue is the source for current enabled models, customer prices and limits. Rates are USD per million tokens, with separate input, cached-input and output rates where available. These are customer charges, not measurements of SingleAPI's provider bill. This guide does not publish a separate price table that can drift from the live catalogue.
The token-cost calculation uses:
(uncached input tokens × input rate
+ cached input tokens × cached-input rate
+ output tokens × output rate) / 1,000,000Settlement uses precise USD nanos and the admitted hold as a cap. Supported hosted tools can add separately published charges. Reasoning tokens, when counted within provider-reported completion tokens, are already part of output usage and must not be added a second time. Provider support and account-tier limits vary by model.
Top-ups buy prepaid credits; inference consumes those credits. Money received from purchases and credits consumed by requests are different metrics. BYOK has zero SingleAPI charge, while the provider may still bill your provider account.
Checkout currently supports one-time prepaid purchases. Subscription billing, refunds, automated tax collection and generated tax invoices are not implemented. The downloadable billing PDF is a usage/balance statement, not a VAT invoice.
Review activity
Open Usage to see request counts, credits used and token totals. Filter by date, model or API key; the page also shows trends and breakdowns by model and key. Token totals can be incomplete when a provider did not report valid usage. For an individual request's status, token counts and charge, use the usage API with its request ID. A BYOK request can show usage while its SingleAPI charge remains zero.
Date ranges use UTC with an inclusive start and exclusive end. Requests rejected before admission do not appear in request history. In-progress, failed and unknown outcomes are distinct; missing token measurements are not zero usage.
For wallet movements, see Billing transactions on Overview. Use Download PDF there for a statement of customer-facing credits, charges and adjustments. Temporary holds and releases are omitted from the PDF statement. For exact usage and billing fields, see the API reference.
Prompt logs and export
Enable content logging on an API key to capture future requests. In Prompt logs, filter by key, date and time, then download the matching logs as an Excel-compatible CSV. The export uses the active filters and redacts content according to the logging policy; it is not an XLSX workbook. See prompt log fields and limits.
Traffic metrics
Usage shows request and token rates, errors, latency and charges from TimescaleDB summaries. Token totals may be incomplete when provider usage is missing. See the traffic API for buckets and coverage fields.