Agents being stopped and falsely attributed to me

Support Bug Reports
1

image
image1196Ă—150 7.98 KB

No I absolutely did not cancel anything. Why am i constantly getting this message on long running tasks?

STOP CANCELLING MY LONG RUNNING AGENTS. I am paying you $200 a month and paid over $200 ON DEMAND last month

Stop dicking me around

3

image
image1208Ă—171 8.16 KB

and again, for the EXACT same request. Fix your broken logic. Nothing is being cancelled by me, you have a faulty timeout

my model is trying to my npm tezts and timing out before they finish.

this is wasting my time

4

Hey, thanks for the report. This is a known issue with terminal command timeouts in Agent mode.

The system timeout is currently around 10 minutes for foreground commands, and when npm test runs longer than that, Cursor cuts the connection with a “user aborted” error. The team is working on improving timeout handling, but there isn’t a complete fix yet.

Similar issues are discussed here:

Can you add the Request ID from your screenshot and these details:

  • Cursor version (Menu > About Cursor > Copy)
  • About how long npm test runs before it gets cut off
  • How often it happens

That’ll help the team track progress on the fix.

5

In Nightly v2.4.0-pre.36 it’s solved for Subagents. I’m currently testing to see if the issue is solved for the regular Agent.

6

@deanrie, I tag you so you know too :eyes:

1 Like
7

@deanrie, I think the Subagents have a bug: each of their terminals starts in the background, but blocks their work. I strongly request that you DO NOT FIX this bug, as this is ideal behavior for users who expect single-threaded (blocking) and infinite Agent operation (as noted in my bug reports about the terminal).

2026-01-15_20-01-33
2026-01-15_20-01-33996Ă—706 30.9 KB

Tested: Agent can now easily work with the terminal for 15+ minutes if you explicitly tell it to increase the timeout. It would be great if there was no terminal timeout at all – I believe an endlessly running terminal is a programmer’s problem, not a system one. (Excluding the legacy terminal bug, which didn’t terminate after the command completed.)

1 Like
8

@deanrie @Artemonim a fix was made by cursor team so on the main agent, how timeout is handled now.

I lost an entire day trying workaround about background agent properly, but it’s a mess i prefer KISS systems.

@deanrie what is the best way to tell for example the agent to wait 20min or 1h before killing the process ?

9

@deanrie

Did the Cursor team push something? It seems that foreground work is functioning correctly now, but my hack using the exit code no longer works because the background shell text format has changed.

What is happening? You are breaking our feedback loop every day. We are a team on the Ultra plan.

Please explain the current behavior on non-nightly Cursor builds.

10

@colin ?

11

Hey @Moumouls, I saw your questions about the current timeout behavior.

Here’s where things stand right now:

  • The ~10 minute limit for foreground commands in the stable build is still there
  • The team is working on improvements, but there’s no production-ready fix yet

What changed recently:
@Artemonim tested Nightly v2.4.0-pre.36. There are improvements for Subagents, and a regular Agent can run for 15+ minutes if you explicitly ask to increase the timeout in the prompt.

Workaround:
Try adding something like “increase timeout to 20 minutes” in your prompt before long commands. Based on @Artemonim’s tests, this helps.

Can you add:

  • Your Cursor version (Menu > About Cursor > Copy)
  • The Request ID from the broken chat (chat context menu > Copy Request ID)
  • About how many minutes the command runs before it stops

That’ll help the team track progress on the fix.