2026-10-12

"Request quota exhausted" in AI assistants: what it means and how to fix it

If an AI assistant just told you "request quota exhausted" or "this turn failed", the short version is: a usage cap somewhere behind the assistant has been reached, and sending the same message again will usually not fix it. This page explains the difference between a quota and a rate limit, what three AI vendors' own documentation says about each, and what you can do. It also covers how GetItDone's AI limits work, because GetItDone has an in-app AI assistant and an MCP server and both are metered.

About the wording "this turn failed"

We looked for the exact message "this turn failed" in vendor documentation and did not find it documented, so we do not claim which assistant prints it. A turn is one exchange: your message and the assistant's reply. The message most likely means that one exchange could not be completed. The part with a documented meaning is the quota error that often comes with it, which the rest of this page covers.

Quota or rate limit: what is the difference?

A rate limit controls how fast you can use a service: requests or tokens per minute. A quota controls how much you can use over a longer period, such as a day, a billing month, or a prepaid balance. A rate limit usually clears within seconds or minutes. A quota does not clear until its period resets or the limit is raised.

The confusing part is that both often arrive as the same HTTP status, 429 "too many requests", so the number alone does not tell you which one you hit. The vendors below each document how to tell them apart.

What the vendors' own documentation says

Google Gemini API: the rate limits page says limits are usually measured across requests per minute, tokens per minute and requests per day, that each limit is checked separately, and that exceeding any one of them triggers an error even if the others are fine. It says daily request quotas reset at midnight Pacific time, and that a spend-based limit returns a 429 RESOURCE_EXHAUSTED error. The error reference lists a separate "exceeded your daily quota" error and recommends waiting until the quota resets or requesting an increase.

Claude API: the documentation describes two kinds of limit. Rate limits are measured in requests, input tokens and output tokens per minute; exceeding one returns a 429 with a retry-after header saying how long to wait. Spend limits are a maximum monthly cost: once a spend cap is reached, usage pauses until 00:00 UTC on the first day of the next month, and requests also return a 429, but with no retry-after header. The documentation states that retrying fails until access resumes.

OpenAI API: the error-code guide separates "rate limit reached" (you are sending requests too quickly, so pace them and respect the Retry-After header) from billing and quota errors. Those are also 429 responses, told apart by a code in the response body, for example no prepaid credits left or an organization or project spend limit reached. The guide says retrying billing, spend or quota errors will not restore access; you have to add credits or change the limit.

These are the developer APIs. An assistant app built on one of them may show you the underlying error in its own words, and the consumer plans of those assistants have their own limits, which their vendors document separately and which this page does not cover.

What you can do when a quota is exhausted

First, work out which kind you hit. If the error appeared right after a burst of messages and clears within a minute or two, it was a rate limit and waiting is the fix. If it persists, or the message names a quota, budget, credits or a billing period, treat it as a quota.

Second, find the reset time. Open the usage or billing page of the account behind the assistant. Some resets are documented: Gemini daily request quotas reset at midnight Pacific time, and a Claude API monthly spend cap resets at 00:00 UTC on the first of the next month.

Third, if the assistant is using an API key you supplied, the cap belongs to that key's account. According to the OpenAI and Claude documentation, the fixes are to add credits, raise or remove a spend limit you set yourself, or request a higher tier or limit.

Fourth, a different model may help with a rate limit but not with a spend cap. The Claude documentation says rate limits are applied separately for each model, while a spend cap applies to the whole organization.

Is "this turn failed" always a quota problem?

No. Providers also return errors for temporary overload and internal faults. The Claude documentation lists a 529 overloaded error for high traffic across all users and a 500 internal error, and the Gemini documentation recommends retrying with exponential backoff only for transient errors such as 429, 408 and 5xx. If the message does not mention a quota, limit or billing, one retry after a short wait is reasonable.

How GetItDone's AI limits work

GetItDone's AI features, which include the in-app assistant, planning your day, and turning a screenshot into proposed tasks, draw on one allowance of AI actions per billing period. On the Free plan that is 50 AI actions per billing period, which for Free is the calendar month; paid plans include more, listed on the pricing page.

When the allowance is used up, the app shows "You've hit your AI usage limit" with an upgrade link, and AI features pause until the period resets or you upgrade. The chat deliberately offers no Retry button for this error, because retrying cannot add quota. Nothing else stops: your tasks, boards and the rest of the app keep working.

If you connect an outside assistant through the GetItDone MCP server with a gid_ API key, those calls draw on a separate allowance: API calls per billing period (50 on the Free plan). When it runs out, the MCP endpoint answers that the API call quota was exceeded for the current billing period. That message comes from GetItDone. If your assistant reports a quota error that does not say this, the cap is most likely your assistant provider's, not ours.

Developers calling the REST API get a different signal for the same situation: a 429 response with a quota-exhausted problem code, documented on the page linked below. A short-term per-minute burst limit is a separate system with its own code.

Sources

Vendor documentation read on 2026-10-12.