Quotas and pricing

Who pays

The user pays, from the one wallet they hold on the platform. Every call that consumes a model debits that wallet, whichever app made it. The platform handles user payments and settlement; apps do not hold user funds.

Rates are set per model and per lane, published in the console and shown inside the product before use. They track upstream cost and can change; the current rate is the one displayed.

What a quota is

Each app has a daily point quota. This limits the points an app can consume each day. It does not add a fee. Usage per app, service and lane is visible in the developer console, together with how much of today’s quota is used.

When a user’s wallet cannot pay, the call returns 402 insufficient_credits. Show the user where to top up (their account page on the platform). When an app reaches its daily quota, requests from that app are rejected until the quota resets; the console shows the reset time. The exact status code for that case is part of the console API, which is still being finalized.

Free lanes

Free lanes do not guarantee a specific provider and may have lower rate limits. They are marked in the catalog and can be used for an initial trial.

Reading cost

prompt-diagnostics reports the cost of the last completed turn and how the context was composed. suggest-reply returns cost and cached directly. Use these to show the player what a turn cost; do not compute prices client-side.