Error messages

Errored, Not Charged

Errored, Not Charged

You'll find this on the Cursor dashboard, in the billing column of your usage events, on the row for a request that went wrong:

Errored, Not Charged

Cursor staff also write it as "Errored, no Charge", and in 2025 there was a sibling, "Aborted, Not Charged", for requests you stopped yourself. As of September 25, 2026 the Cursor forum has 24 topics that mention the phrase, nearly all asking the same thing: was I charged or not?

What the label means

The request errored before it finished, and Cursor didn't bill it. A staff member put the rule plainly in June 2026: "You won't be charged for requests that fail to generate a response." In the same reply the Usage tab's values were listed as Free (a promotion, or a request that "errors out before completing successfully"), Included (covered by your plan) and a dollar amount for on-demand usage. So a failed request can also appear as Free, with no tokens counted. Cursor changed that token cell to show "-" instead of 0 in June.

Staff said the same about requests that failed during a service issue on September 17, 2026: they "were rejected before model usage began, so you were not charged".

Why an errored request shows as charged at first

This is the usual complaint, and it has an answer. The dashboard records the request first and settles the bill afterwards. In January 2026 staff wrote that "there can be a delay between when the request first shows up in the dashboard, and when it gets reconciled as 'Errored, no Charge'." In March the advice was the same for provider errors, which "can show up in usage at first. Usually, the system fixes it on its own after a while". Another reply mentioned that the usage data is aggregated hourly, so a fresh row can lag too.

• Wait and reload the usage page before assuming you paid.
• If it still shows as charged, email hi@cursor.com. Staff said the team "can review it and correct the usage on your account."
• Keep the Request ID (chat menu, Copy Request ID), since that's how staff trace a specific request.

Stopping a request yourself isn't free anymore

This is the part that caught people out. Until mid-October 2025, a response you stopped was marked as not charged. From about October 15, 2025 stopped requests started being billed like any other, and a 16-post thread asked whether that was a bug. Cursor staff said on October 22 it was "an intentional change. We started charging for partial requests since they can be used to bypass the system and still consume resources." A forum regular with the Community Developer title (not marked as Cursor staff) added that a stopped request "was already submitted to the AI provider who does charge for the consumption."

So the free label is for requests that fail on Cursor's or the provider's side. If you stop the agent because it's heading the wrong way, expect to pay for the tokens it used up to then. We'd assumed stop and error were treated alike, and they used to be.

When "Errored" is a setup problem

Errored rows sometimes come from your own configuration. In April 2026 staff tied "Errored, No Charge" on a user's Azure-hosted model to a known bug with Azure OpenAI personal keys, where Cursor wasn't passing the model name to Azure. If every request to one model errors, check Settings > Models > API Keys for a personal key that's switched on. You won't pay for those failures (small comfort when nothing gets answered).