[BUG] VS Code: "Not responding - try stopping" offers no way to actually stop or recover
What's Wrong?
When Claude Code gets stuck during a task, the UI shows "Not responding - try stopping" — but there is no actionable way to stop. No stop button, no cancel, no retry. The only option is to kill the session entirely and lose all progress.
This is the worst UX failure in Claude Code right now. "Try stopping" is not a solution — it's the app throwing its hands up. The user is left staring at a dead session with no recourse.
What Should Happen?
When a task stalls, the extension should fail gracefully:
- Show a visible Stop/Cancel button next to the "Not responding" message — not just text that says "try stopping"
- Allow recovery without killing the session — skip the stuck task and continue, or retry it
- Timeout and fail the individual task after a reasonable period (30-60 seconds) rather than hanging indefinitely
- Preserve completed work — if 7 of 10 tasks completed successfully, those results should survive even if task 8 fails
Right now "Not responding" is a dead end. The user has to force-quit and restart from scratch.
Steps to Reproduce
- Open Claude Code in VS Code (extension v2.1.84)
- Start a multi-step task that uses the todo/task panel (e.g., building a new repo with ~10 steps)
- Let Claude work through several tasks (7 of 10 completed in this case)
- When a task involves an external operation (like
gh repo createor a network call), it stalls - The task shows a red asterisk. The bottom of the panel shows:
* Not responding · try stopping - There is no stop button, no cancel, no skip. The session is dead.
Error Messages/Logs
No error message — just the red text:
* Not responding · try stopping
No stack trace, no timeout error, no indication of what went wrong.
Environment
| Detail | Value |
|--------|-------|
| Claude Code Version | 2.1.84 |
| Claude Model | claude-opus-4-6 (1M context) |
| Platform | Max plan |
| Operating System | macOS 26.3.1 (Darwin 25.3.0) |
| Terminal/Shell | VS Code integrated terminal (zsh) |
| Machine | MacBook Pro, Apple M2 Pro, 16 GB |
| VS Code | 1.112.0 |
| Claude Code Extension | anthropic.claude-code@2.1.84 |
Is this a regression?
Don't know — but this happens frequently enough to be the single most frustrating part of using Claude Code daily.
Additional Information
The core problem is that "Not responding" is treated as informational text rather than an error state with recovery options. Compare to how VS Code itself handles unresponsive extensions — it offers "Restart Extension Host" as a clickable action. Claude Code should offer equivalent escape hatches.
Related issues:
- #39229 — extension host heartbeat timeout (same symptom, different root cause)
- #39183 — "Not responding" on WSL2
- #38580 — docs missing the red spinner indicator
This is not an edge case. It happens regularly during multi-step tasks and forces a full session restart every time.
73 Comments
Im experiencing this now...
i'm also experiencing that but when waiting enough time, it works for me
I have been experiencing this issue too. The first couple of times I had this happen to me, I was able to click the stop button (what was the submit button), and then I asked Claude to continue what I was previously working on. Usually, it would say that the work was already completed. In my experience, this issue seems to occur when I'm in Planning mode. I cannot say it only happens in Planning mode.
After experiencing this a third or fourth time, I searched to see if this bug had been reported as an issue. In the time it took me to find this Issue (maybe 2-3 minutes at most), I went back to my VS Code window and saw that the "Not responding - try stopping" message was gone and I had another permission prompt waiting for me.
This is definitely an inconvenience. If I continue to experience this issue, I'll start timing how long it takes for it to recover.
I have been experiencing this issue too
Damn it started with me now as well
the same issue
The same issue here, and happens in all modes
Same issue, and I've to kill it
Same issue, it takes a large amout of time, but it works
I'm going through the same thing. Between today and yesterday I experienced this at least 5 times, and in one instance it consumed almost 10% of my session usage.
Same issue. Does anyone have a fix ? Do we just wait it out ??
same issue, that's so annoying and it seems like a new issue, never seen that before
actually to fix that u just need to wait around 2-3 minutes, for me it's gone
Disaster! I have the same problem! - It's frustrating!
Same issue here! I've been experiencing it on many occasions over the last 20 hours.
I'm completely blocked due to this. Damn!
Same issue. restarting again and again eating up the limit
I am experiencing the same. Very annoying.
Temporary workaround: when using the CLI and not the Vscode plugin, I have no issues.
Just to add to the chorus of folks suffering this: I've been fighting this same issue all day today. Didn't see it ONCE yesterday.
I'm not sure if this will help anyone, but I found this in the release notes for v2.1.83:
So apparently it's a feature....?
I've had a few cases where it was just waiting on output and continued after it got the expected results.
Same message has brought me here :) I did notice the extension chat content was at 98% and going to compact soon. Each time this has occurred for me, I end up disconnecting the remote explorer connection in VS code and reconnected, which seems to work. I try to copy the last few chat lines to remind Claude where we left off.
I asked claude about this issue, and the chat said it occurred when the backend was processing and the front end stalls (or something like that).
I am experiencing it too, it is definitely a very recent thing it started for me, with the Claude blackout a couple days ago, i thought the issue was my server dev environment, but it is happening in all my local machines too.
killer dude this actually make sense haha!
2026-03-26 23:21:04.331 [info] Received message from webview: {"type":"request","channelId":"njpkggvknrk","requestId":"ma7yiqy9z3k","request":{"type":"log_event","eventName":"ide_backend_stall_detected","eventData":{"idleMs":60010}}}
The session started, the webview sent the request. The IDE backend accepted the task, but could not proceed, probably due to a resource lock (deadlock) or an infinite indexing loop. Since nothing happened for 60 seconds (neither a stray token nor a filesystem message), the system generated a log_event about the stall and threw a “Not responding” message to the UI, also under git init. LoL.
Fixing this in tomorrow's release! The timeout we set there was a bit too aggressive.
Same! Left it for a few 30 or so seconds then it carried on 👍
This is frustrating, is there any way to recover state and continue while waiting for a fix?
Claude desktop pointed me to this bug, and suggested I restart the extension host, but that results in window not responding, reload window, same as restarting vscode and claude does not remember what it was working on.
P.s. It also has the cheek to start making changes while in plan mode when this stopped responding issue starts working again by itself, and when I give it a screenshot of plan mode being active, it suggested I must have changed the mode after it started working.
Same issue here
I think you can manually install v2.1.82 for the extension until tomorrow's update fixes it.
Same issue, and so far unrecoverable. Never continues after waiting; must kill the session and restart, and often that doesn't work because it kicks off the new session with the same multi-step process that caused the issue in the first place.
Hopefully, this has now been fixed with the recently released version 2.1.85.
also experiencing this a ton today. noticed it for the first time today, even though my usage patterns haven't been any different. The most frustrating is when it happens during a /compact. Then I stop and restart the /compact and maybe it happens again. How many tokens am I wasting on these cycles that lead to nothing??
I did update to 2.1.85 .... still happening.
Same issue here
Yep same issue here. Just leave it for a bit and it will fix it self. As others have stated below
same issues here
I've had this happen to me at least 5 times to me today, even when i restart my session from scratch and ask it a different task that is easy to do it continues to give me this answer
Same issue right now.
so far Opus only bug for me
same issue, pretty annoying, and can happen with pretty much any prompt, from simple to complex
Same issue here in 2.1.85
zelfde probleem al paar dagen lang, updaten heeft niet geholpen
Same issue here in 2.1.85. Usage data isn't even loading.
<img width="474" height="102" alt="Image" src="https://github.com/user-attachments/assets/b8d4280c-ae0a-4c45-9787-a3bb9984791d" />
Same problem, I emptied the %APPDATA%\Code\User\workspaceStorage folder and reverted to version 2.1.81. The message is gone, but Claude Code isn't processing the prompts, the "Thinking" section isn't appearing.
facing same issue here from past 2 hours
same issue nothing works
The problem stems from the Opus model. It continues to work if you switch to Sonnet. You can proceed this way until it's resolved :)
Same issue here. Sometimes pressing stop works after a while and then continue. Otherwise usually just quitting VSCode and a restart helps.
@qing-ant ??
Same thing here. Eventually catches up and resolves itself but very frustrating.
Same issue, but after waiting for around 2-3 mins, it starts back.
<img width="292" height="78" alt="Image" src="https://github.com/user-attachments/assets/e1e5df3b-8561-4d85-afc4-f1110758cadf" />
Current workaround: break the task into smaller steps and ask the agent to execute them one by one instead of giving it a large chunk all at once.
Can anybody please explain how is it starting after 2-3 mins , I have been waiting for more than an hour and still no luck, tried restarting VS Code as well , but no luck
same error here
using claude code in the terminal fixed it for me, for now at least.
very frustating... claude team please look into it.
The same problem for me started today, and it occurred when I used the remote-control to issue commands remotely. There was a question from Claude that required a user selection and I didn't check it for a few hoiurs. When I clicked on it, I did so on my PC in VS Code but previously I had been using my android app to issue prompts.
I was able to fix it by stopping it, converting to plan mode, askign it to summairse what we have been discussing, then go away for 5 minutes and come back. It offered a new user decision which I clicked on and then eveyrhting seemed to go back to normal for me.
I am not sure what is causing this issue, I have not seen it before today.
WORKAROUND FOR CLAUDE CODE STUCK / LOOP ISSUE
If your session is stuck, looping, or refusing to stop, this approach works much more reliably:
Step 1: Ask Claude Code to create a todo.md file for your task
It should break the entire task into clear, smaller steps
Step 2: Do NOT ask it to execute everything at once
Instead, instruct it to implement one step at a time from the todo list
Claude Code tends to hang or loop when given large, multi-step tasks in a single prompt
Using a todo.md + step-by-step execution keeps it focused and prevents the session from getting stuck
Not true. This is happening to me consistently today exclusively in Sonnet.
The question I have is: why isn't it retrying on its own? This can't be hard to add.
It happens every 5 minutes now. Easily becoming the buggiest vs-code extension there is.
even with small tasks clean prompting and a using the tips provided in this thread I keep getting stuck. I really hope anthropic compensates max users with extended usage cause I have not been able to do almost anything in the past 48 hours.
Faced the same problem. However, leave it out there. Don't stop I see that the message goes away and it continues from where it left off
Seems like everyone has been experiencing this the past few days. I just got hit with it too
i noticed this stopped happening to me when i moved away from VSCode and
into Terminal. I've been cooking in terminal for 5+ hours and no issues
It's just a time-out problem.
Just wait and you'll see the session following its course withou issues.
Anyone found the solution?
If you just let it run it fixes itself after sometime. Seems to be isolated to using planning mode
What works for me is this, when it's stuck I type "??" and send, nothing happens, then I hit stop it shows "Interrupted" and then it resumes working.
Still experiencing this..
Same issue here. It now consumes a lot of tokens without generating anything. I’m not sure whether this happens only in the VS Code extension, or if it also occurs when using the command line.
Using the VS code extension this still happens, but if you let it run without stopping it resumes working automatically after a while.
It seems like this is just an aggressive notification in case it actually _is_ stuck. But if it's doing a big file rewrite or something you expect to take a while anyway, then just wait a while and it'll continue.
Either way, it's bad UX, and really the problem is that the VSCode extension has no way to show the tokens it's consuming in realtime, so you can't tell if it's really stuck or just busy cooking.
This issue has been pestering me all morning. I found that toggling my prompt mode helped with getting it unstuck on the error message for "Not responding - try stopping". Not an ideal solution if you let your guard down with the mode set to Edit Automatically. This should work if you're actively stuck and tired of waiting for it to recover.
its so weird, if i click the stop button once, it continues? its so odd, what on earth is this UX glitch lol
This issue has been automatically locked since it was closed and has not had any activity for 7 days. If you're experiencing a similar issue, please file a new issue and reference this one if it's relevant.