[BUG] Terminal tool gets stuck after exit code

Some commands work, some not. No idea why:

1 Like

This is exactly what it’s doing to me! it makes it not even usable. It’s cost me almost a week of my project.

1 Like

I’m curious if anyone from Cursor devs is following this thread. Looks like it’s quite annoying issue for multiple people out there. And I wonder how many aren’t just reporting.

4 Likes

Same issue here, really slowing me down

1 Like

This issue also affected me for the past 2 weeks, up until today when now it’s more severe: the tool shows no output at all. Clean reinstall, tried multiple shells. I have no workaround so far.

1 Like

When I am using claude sonnet 3.7 max, it does not happen. Why different exist ?

Im facing the same issue

1 Like

Popping the terminal gets the agent going again, but it frequently can’t see the results, so it tries the command again (or worse, ā€œa different approachā€) and then gets stuck in the same place. SO FRUSTRATING.

1 Like

I have shell integration enabled and still get this issue. I’m using remote WSL. The issue I have is that it works fine… sometimes. And then it starts hanging and hangs most of the time.

1 Like

+1 exactly as described

1 Like

Please, Cursor team, take a look at this. I can help reproduce or w.e., DM me if needed. This is really a blocking bug in a lot of ways. @danperks

1 Like

+1 exactly as described

Experiencing the same issue. yolo mode worked for me when first starting to use cursor a couple weeks ago, now it’s broken as described.

This is not a fix

Hey there! Yash from the Cursor team here. I am looking into fixing this issue now. Would love to hear from y’all if there are consistent steps you use to reproduce this broken state.

2 Likes

exactly the same issue here.

Env:
Version: 0.48.6 (user setup)
VSCode Version: 1.96.2
Commit: 1649e229afdef8fd1d18ea173f063563f1e722e0
Date: 2025-03-31T05:21:50.407Z
Electron: 34.3.4
Chromium: 132.0.6834.210
Node.js: 20.18.3
V8: 13.2.152.41-electron.0
OS: Windows_NT x64 10.0.19044

Tried solution above but did not help.

I am connect via SSH and simplely ask agent to execute ls.
I could see the command executed and result on ternimal.
the terminal could work after command.
I could input more command in the same terminal with correct result.
Agent is waiting for the result somehow.

I do not know if it’s related.
If I run ā€œcd PATH && make clean && make &&./mainā€, i failed
If ask agent to run command one by one, it’s likely okay…

agree, its not fixed - happe 0.48.6 with long terminal output for me

It looks like its related to the output of a different commands. If the output uses less it will freeze. In my case, I’m bypassing this issue by piping the commands to cat. So something like: Use git diff --staged | cat makes it work for me.

How do you see the log that sheds the light on how the tools are called, terminal tool specifically? I’m seeing so much ā€œtool call timeoutā€ and stuck terminals, I’m inches away from switching to another IDE at this point.

As for ways to replicate, it for sure happens all the time on my Ubuntu machine when I use it via SSH. Consistently around 5 times in an hour, but in some chats it happens once and then repeats forever. Can only tell this much without any transparency in the tool calling process.

Please consider making this more transparent in the UI, less ā€œmagicalā€, at least give an option. I would happily tweak something if I can figure out myself at least how not to trigger bugs like these.

1 Like