Skip to content

Limits & Quotas

Gonka Broker enforces safety limits to protect you from accidental overspending, runaway scripts, and other unexpected situations. These limits are designed to keep you safe, not to restrict normal usage.

Every top-up is checked against three safety limits:

  • Minimum per payment: prevents accidentally tiny charges that would mostly go to payment processing fees.
  • Maximum per payment: prevents typos like adding an extra zero and paying $10,000 instead of $1,000.
  • Maximum per 24 hours: caps total spending within a rolling 24-hour window, protecting against payment loops or other unexpected behavior.

These limits are system-wide defaults. If a payment exceeds any of them, it will be declined and you will see an error message with the specific limit.

There is a maximum number of API keys per account. This encourages good key hygiene: if you need a fresh key, rotate your existing one instead of creating a new key.

The default limit applies to all accounts. If you genuinely need more keys, contact us at info@gonkabroker.com and we can adjust the limit for your account.

Each API key carries its own limits, editable in the dashboard: open the key menu and choose Limits. Three controls are available:

  • Rate limit: requests per minute for this key.
  • Allowed models: which models the key may call.
  • Spending limit: how much the key may spend per day or per month.

This is useful when one account issues keys to different people, projects, or agents: each key gets its own speed, model set, and budget. You can also define default limits for new keys in Settings; they are applied to a key when it is created and never change existing keys.

API requests are rate-limited on a per-key basis (requests per minute). You can set the RPM for each key in the Limits dialog, from 1 up to a ceiling shown in the dialog. If you need a higher ceiling, contact us at info@gonkabroker.com; the ceiling is raised for your account and covers all your keys.

Rejected requests count toward the rate limit: a request refused before reaching a model (an unknown or not-allowed model name, an invalid payload) still uses up a slot.

You can check the current RPM limit for your key by calling the GET /test-auth endpoint:

Terminal window
curl https://proxy.gonkabroker.com/test-auth \
-H "Authorization: Bearer YOUR_API_KEY"

The response includes your key’s maxRpm value in the settings object.

By default a key may call every available model. Setting an allow-list restricts it to the models you pick. A request for any other model returns 404 model_not_found, the same error an unknown model gets. New models added to the platform are not added to existing allow-lists automatically: a restricted key stays restricted to exactly what you chose.

A spending limit caps how much a key spends per UTC calendar day or month, in USD. Two reactions are available when the limit is reached:

  • Notify only: you get an email, the key keeps working.
  • Notify and block: you get an email and the key stops until the period resets at 00:00 UTC.

Either way you also get a warning email when the key passes 80% of its limit. The warning email can be turned off in Settings; the limit-reached email is part of the limit itself and always sends.

Spending limits are approximate: a small overshoot of up to a few minutes of traffic at the key’s rate is possible before the limit takes effect. A blocked key returns 429 with the code insufficient_quota and a message naming the reset time:

{
"error": {
"message": "API key spending limit reached. Resets 2026-01-02 00:00 UTC.",
"type": "insufficient_quota",
"param": null,
"code": "insufficient_quota"
}
}

Raising or removing the limit unblocks the key, usually within a minute.

Spending limits act on paid usage. If a free tier is available and your account is running on free-tier funds only, the free-tier rate limit governs the key instead.

If a free tier is available on your account and it is running on free-tier funds only, requests are capped at a lower rate; the current cap is visible as maxRpm in the /test-auth response above. Hitting the cap returns 429 with a rate_limit_error message explaining the limit. The cap is lifted within about a minute after a top-up. See Billing & Deposits for how the free tier works.

With no remaining balance at all, the API stops accepting new requests and returns 429 with the code insufficient_quota:

{
"error": {
"message": "Balance exhausted. Top up to continue: ...",
"type": "insufficient_quota",
"param": null,
"code": "insufficient_quota"
}
}

Top up to resume; the block is lifted within about a minute.