Usage

Usage shows how many proxy requests your plan has counted in the current window and when that window resets. Open it from the account menu: select your account at the bottom of the sidebar, then Usage.

What is on the page

• Your usage, with your plan on the right: Free or Paid. A paid plan that is not in good standing carries a badge such as Past due, Paused or Canceled.
• The allowance row. On Free it is headed Proxy requests this month and shows the count with Unlimited in place of a meter. On a paid plan it is headed Proxy requests this period and shows used of the plan's allowance, a meter and the percentage used.
• In flight: requests that have started but not finished. They are reserved against the allowance and settle when they end.
• When the count starts over: Resets on Free, Renews on a paid plan, or Access ends on a paid plan you have canceled.
• Last updated: when the numbers were fetched. The refresh button in the header fetches them again.
• Manage billing: appears only when you have a paid subscription. It opens the payment provider's billing portal, where you manage or cancel the subscription.

Your plan

Every account starts on Free. Free has no request ceiling. Requests are still counted, per UTC calendar month, so the page can show them, and the count starts over at the start of each month (UTC).

A paid plan has a fixed number of requests per billing period, and the count starts over when the period renews. When the allowance is used up, further requests are refused with HTTP 429 until the period renews. When paid access ends, you go back to Free for the rest of that calendar month; the Free requests you made earlier that month are still counted. Paid checkout is not open yet; see Upgrade plan.

What counts as a request

One call to a model through the proxy counts as one request, whatever its size. That covers chat completions, Responses, Messages, audio transcriptions and token counting, on the shared endpoints and on group endpoints.

These do not count:

• Listing models and every request the app itself makes.
• A request rejected for a bad key, a bad body or an unknown model.
• A request that fails upstream before any output reaches you. A stream counts once it has delivered content or tool output, or finished normally.

Tokens and estimated cost are on Overview and Logs, not here. Limits a team owner sets for you are on Team and are checked after this allowance.

After a payment

When you come back from checkout, the page says your payment is being confirmed. The subscription appears once the payment provider confirms it, which can take a moment. If it takes longer, the page says so; choose Refresh. You do not need to pay again.

Related pages

• Upgrade plan: the paid plans.
• Team: limits a team owner gives you on shared accounts.
• Overview: tokens and estimated spend over time.