[BUG] Cowork on Windows 11 — VM crashes 6-12 times daily requiring manual service restart, plus false rate limit errors requiring new session
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?
Cowork on Windows 11 is critically unstable for sustained daily use. The CoworkVMService crashes repeatedly throughout the day — typically 6-12 times — requiring a manual Windows service restart and application relaunch each time to recover.
Separately, a false "rate limit reached" error appears in the Cowork window with increasing frequency. This is not related to actual usage (I am a Max plan subscriber well within my limits). Once it appears, nothing within the session clears it. The only workaround is to open an entirely new Cowork window using the same folder. This is happening more frequently over time, now occurring close to once a day.
Both issues have been worsening progressively over recent weeks.
What Should Happen?
Cowork should maintain a stable VM connection throughout a working session without requiring manual service restarts. The rate limit error should not appear when usage limits have not been reached. If it is legitimately triggered, it should clear automatically and display an accurate message.
Error Messages/Logs
- "API Error: Rate limit reached" displayed in the Cowork chat window despite being well within Max plan usage limits. Cannot be dismissed or cleared within the existing session.
- CoworkVMService crashes silently — the Cowork interface becomes unresponsive with no error message displayed in the UI.
Steps to Reproduce
Both issues occur during normal daily use with no specific trigger — they are stability failures rather than reproducible sequences.
Current recovery procedure for VM crashes:
- Open Windows Services (services.msc)
- Stop CoworkVMService
- Start CoworkVMService
- Relaunch Claude Desktop
An alternative workaround (clearing VM cache and restarting vmcompute and vmms via PowerShell as Administrator) also resolves the crash state temporarily but does not prevent recurrence.
The rate limit error has no recovery procedure within the existing session. Opening a new Cowork window pointed at the same folder is the only workaround.
Claude Model
Not sure / Multiple models
Is this a regression?
No, this never worked
Last Working Version
_No response_
Claude Code Version
Claude 1.1.5368 (d12d02) 2026-03-05T17:29:33.000Z
Platform
Anthropic API
Operating System
Windows
Terminal/Shell
Other
Additional Information
I am a Max plan subscriber. These issues make Cowork non-functional for sustained professional use. I am currently creating up to a dozen new Cowork sessions daily due to VM crashes, and opening an entirely new Cowork window almost daily due to the false rate limit error.
Both issues are worsening over time, suggesting a regression rather than a static bug.
Related issues: #29848, #28094, #25516, #25505, #28793
This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗