Where do I report a Cursor bug?

The Cursor forum at forum.cursor.com is the documented route, and it gets community help and visibility as well as staff attention. Account and billing problems go to support by email instead, since those are not bugs.

What is a Cursor request ID?

A unique identifier generated for each request you make to Cursor. It lets support find your specific request in their internal systems and see what went wrong. Without one, a report about a single bad response is very hard to act on.

How do I find my request ID in Cursor?

Open the conversation in the Chat sidebar, click the context menu shown as three dots, and choose Copy Request ID. Then paste it into your forum post or email.

Is a request ID sensitive? Should I redact it?

No. Cursor states that request IDs are only meaningful inside its own backend and are lookup keys with no value elsewhere, so you do not need to treat them as confidential. Posting one publicly on the forum is the intended use.

Why can't support see what my agent did?

Almost always Privacy Mode. With it on, support can see which model was used, whether tool failures occurred, and backend failures unrelated to your work. They cannot see which tools failed, the conversation, or the context the agent had. That is enough to debug a connection problem and not enough to debug agent behaviour.

Does turning off Privacy Mode help with a bug that already happened?

No. Privacy Mode is applied per request, and each request is logged according to the setting in force when it was submitted. Changing the setting now does nothing for yesterday's request. You have to reproduce the problem with Share Data on to produce a request support can actually read.

How do I report an agent doing something wrong?

Enable Share Data, reproduce the problem, copy the new request ID, send that one, then switch back if you prefer. The first request ID, captured under Privacy Mode, is usually not worth sending.

What if the problem is a connection or sign-in failure rather than the agent?

Those are often debuggable with Privacy Mode still on, because they do not require seeing your prompt or code. Send the request ID, your Cursor version and the operating system, and check the status page first in case it is already known.

Reporting a bug in Cursor

Most reports stall for one reason: no request ID, or a request ID recorded under Privacy Mode, which shows support almost nothing.

forum.cursor.com
Request ID required
Privacy Mode is per request

Checked against cursor.com on 25 September 2026

The one thing to get right

A request ID is a unique identifier generated for every request Cursor makes on your behalf. It is how support locates your exact request among everyone else’s and sees what happened to it. Copy it from the conversation’s context menu, the three-dot menu in the Chat sidebar, using Copy Request ID.

You do not need to redact it. Cursor says request IDs are only meaningful inside its own backend and have no value outside those systems, so pasting one into a public forum thread is exactly what it is for.

What the request ID reveals to support, though, depends entirely on a setting you probably configured months ago and have not thought about since.

What support can see

Your privacy setting decides what is attached to the request they look up. This is the difference between a fixable report and a thread that goes nowhere.

Visible to support	Privacy Mode	Share Data
Which model was used	Yes	Yes
Whether tool failures occurred	Yes	Yes
Which tools failed	No	Yes
Backend failures unrelated to your prompt or code	Yes	Yes
The full conversation with the agent	No	Yes
Context given to the agent: system prompt, rules, git status	No	Yes

Read the middle rows again. Under Privacy Mode, support can tell that a tool call failed but not which one. For a connectivity problem that is usually enough. For “the agent deleted my migrations” it is nothing at all.

Reporting unexpected agent behaviour

1Temporarily enable Share Data in your privacy settings.
2Reproduce the issue by doing the same thing again.
3Copy the new request ID from the conversation's context menu.
4Send that request ID to the forum or to support.
5Switch back to Privacy Mode if that is your preference.

Step two is the one people skip, and it is the one that makes the report useful. A report that says “it did this yesterday” with a Privacy Mode request ID attached gives support a model name and nothing else.

The rest of the report

Cursor’s own checklist, with where each item comes from. Collecting all eight takes about two minutes and saves a round trip that takes a day.

Include	Where to find it
Cursor version	Cursor > About Cursor in the menu bar.
Operating system and version	Mac, Windows or Linux, with the release number.
Steps to reproduce	What you did immediately before it happened, in order.
Expected behaviour	What you thought would happen.
Actual behaviour	What happened instead.
Screenshots or a recording	Only needed when the problem is visual, but it usually shortens the thread.
Request ID	Copy Request ID from the conversation's context menu.
Console errors	Help > Toggle Developer Tools, then read the console.

Console errors are worth the extra step when the editor itself is misbehaving rather than the model: a window that will not load, an extension that breaks the sidebar, a crash on startup. Help, then Toggle Developer Tools, then copy whatever is red.

Where to send it

Post on forum.cursor.com. That is the documented destination for bugs, and the public thread means someone hitting the same problem next week finds your post instead of filing a second one.

Before you post, check status.cursor.com. A large share of reports that feel like bugs are an ongoing incident, and the status page answers that in five seconds.

Things that are not bugs go elsewhere: billing, refunds, a subscription that will not cancel and account access all go to Cursor support, which can see your account. The forum cannot.

Checked against Cursor’s bug reporting guide on 25 September 2026.

Frequently asked questions

Where do I report a Cursor bug?: The Cursor forum at forum.cursor.com is the documented route, and it gets community help and visibility as well as staff attention. Account and billing problems go to support by email instead, since those are not bugs.
What is a Cursor request ID?: A unique identifier generated for each request you make to Cursor. It lets support find your specific request in their internal systems and see what went wrong. Without one, a report about a single bad response is very hard to act on.
How do I find my request ID in Cursor?: Open the conversation in the Chat sidebar, click the context menu shown as three dots, and choose Copy Request ID. Then paste it into your forum post or email.
Is a request ID sensitive? Should I redact it?: No. Cursor states that request IDs are only meaningful inside its own backend and are lookup keys with no value elsewhere, so you do not need to treat them as confidential. Posting one publicly on the forum is the intended use.
Why can't support see what my agent did?: Almost always Privacy Mode. With it on, support can see which model was used, whether tool failures occurred, and backend failures unrelated to your work. They cannot see which tools failed, the conversation, or the context the agent had. That is enough to debug a connection problem and not enough to debug agent behaviour.
Does turning off Privacy Mode help with a bug that already happened?: No. Privacy Mode is applied per request, and each request is logged according to the setting in force when it was submitted. Changing the setting now does nothing for yesterday's request. You have to reproduce the problem with Share Data on to produce a request support can actually read.
How do I report an agent doing something wrong?: Enable Share Data, reproduce the problem, copy the new request ID, send that one, then switch back if you prefer. The first request ID, captured under Privacy Mode, is usually not worth sending.
What if the problem is a connection or sign-in failure rather than the agent?: Those are often debuggable with Privacy Mode still on, because they do not require seeing your prompt or code. Send the request ID, your Cursor version and the operating system, and check the status page first in case it is already known.