[BUG] Claude Code in Claude for Mac fully processes request, but stops providing output before completion
Preflight Checklist
- [x] I have searched existing issues and this hasn't been reported yet
- [x] This is a single bug report (please file separate reports for different bugs)
- [x] I am using the latest version of Claude Code
What's Wrong?
With increasing frequency, Claude Code in the Claude for Mac environment receives and processes a request, but stops providing output before reaching completion.
When asked if it completed its task, it responds that it did, and often refers to output that is not displayed in the UI.
Example #1:
<img width="708" height="230" alt="Image" src="https://github.com/user-attachments/assets/cd96a2ba-0275-4736-a2e1-d99f7dbb3097" />
Example #2:
<img width="740" height="673" alt="Image" src="https://github.com/user-attachments/assets/d310be80-33c4-4b80-a113-a25366c9ea6c" />
To be clear - this is not a matter of user impatience. Claude Code starts processing the task, initially provides output, and then stops providing output indefinitely - 30 minutes, 60 minutes, overnight, etc. A follow-up prompt results in an immediate answer from the agent, which often repeats output that the agent apparently believes that the chat should have shown.
Sometimes, the missing output is retroactively inserted into the chat before my request as to whether or not my request was satisfied. But it does not appear until and unless I submit my request, even if I restart Claude for Mac.
Claude Code Desktop has previously exhibited this behavior, but it is getting worse and occurring more frequently over time.
What Should Happen?
Claude Code in Claude for Mac should continue to provide output through the end of processing a task without having to be nudged.
Error Messages/Logs
Steps to Reproduce
- Make a request of Claude Code in Claude for Mac.
- Watch for output. Occasionally notice that output from Claude Code stops before the task is complete.
- Ask Claude Code if it is still processing the request. Notice that it immediately responds that it did, and often refers to output that was not previously shown.
Claude Model
Opus
Is this a regression?
Yes, this worked in a previous version
Last Working Version
Some version before 1.1.4173 (12766c)
Claude Code Version
1.1.4173 (12766c)
Platform
Anthropic API
Operating System
macOS
Terminal/Shell
Other
Additional Information
_No response_
8 Comments
Found 3 possible duplicate issues:
This issue will be automatically closed as a duplicate in 3 days.
🤖 Generated with Claude Code
Not a duplicate. There is no "stop_reason" shown, as per issue #1; and the behavior does not involve text that is initially shown and then removed, as per issue #2 and issue #3. This is different behavior that requires attention.
This is getting much worse. Every other request I submit ends with output being truncated, and I have to nudge it to see what happened with my request, whether it is done processing, or whether it even started processing.
<img width="758" height="628" alt="Image" src="https://github.com/user-attachments/assets/16f5ddea-efaf-4509-a312-e0baa46cecb3" />
This continues happening! In probably 10-20% of requests that I issue to Claude Code through Claude for Mac, the chat session just tapers off and stops responding, no matter how long I wait. And when I ask for a status update, it immediately says that it's done and refers to messages that did not appear in the chat. Again, as described above. And sometimes, the missing messages are backfilled, being inserted into the chat immediately before my status-check request, but after I submitted the request.
This is very irritating and requires attention. I have also submitted this as a bug in the Claude Code Desktop app.
<img width="747" height="324" alt="Image" src="https://github.com/user-attachments/assets/fb1a0a8a-6a7c-4328-9c84-4ecaf2770a6c" />
This continues to occur with high frequency.
<img width="732" height="489" alt="Image" src="https://github.com/user-attachments/assets/74b828d9-53fa-4b23-b545-9192bf1f06ba" />
Closing for now — inactive for too long. Please open a new issue if this is still relevant.
I want you to know that I canceled two x20 MAX subscriptions because this issue was never resolved, and I instead subscribed to Codex Pro. My $200 is better spent on agentic coding solutions that do not exhibit narcolepsy, and companies that actually care about their user experience.
I will not be using Anthropic for any purpose at any point in the future.
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.