claude-code - 💡(How to fix) Fix [BUG] CoworkVMService stops silently on Windows — app crash-loops with no recovery path [1 comments, 2 participants]

Bug Description

The CoworkVMService Windows service stops on its own during normal use. When this happens, every attempt to open Cowork triggers a crash-restart cycle — the app fires cowork-vm-shutdown, which fails ("VM service not running"), then Sentry errors fire, and the app relaunches. This loops indefinitely with no recovery.

The service is set to StartType: Automatic but does not auto-restart after stopping. The only fix is manually running Start-Service CoworkVMService in an elevated PowerShell — but there is no UI indication that this is the problem.

Related

• #43534 — same user, different failure mode (stale VM survives Windows Fast Startup)
• #29941 — CoworkVMService fails to start due to signature verification
• #27801 — "Failed to start Claude's workspace" — VM service not running

Log Evidence

Apr 9: First occurrence — 5 crash cycles + 30 min of continuous failures

The service died at ~19:26. The app crashed and restarted 5 times in 19 minutes, then spent 30 minutes continuously failing to boot the VM:

19:16:47 [info] Successfully run onQuitCleanup: cowork-vm-shutdown  ← last success
19:26:45 [error] Sentry caught (x5)  ← crash-restart #1
19:36:09 [error] Sentry caught (x5)  ← crash-restart #2
19:38:09 [error] Sentry caught (x4)  ← crash-restart #3
19:41:09 [error] Sentry caught (x5)  ← crash-restart #4
19:45:02 [error] Sentry caught (x5)  ← crash-restart #5
19:51:12 [error] Failed to run onQuitCleanup(cowork-vm-shutdown): Error: VM service not running. The service failed to start.
20:02:20 [warn] [HostLoop] VM boot failed; bash proxy unavailable: VM service not running.
20:15:49 – 20:31:41  ← 11 more VM boot failures in succession
21:01:05 [error] Still failing at end of session

Apr 12: Two separate incidents (morning + evening)

Morning:

10:09:17 [error] Failed to run onQuitCleanup(cowork-vm-shutdown): VM service not running.
10:11:15 [error] Session initialization failed: VM service not running.
10:13:05 [error] Sentry caught (x7)

Evening:

20:38:40 [error] Failed to run onQuitCleanup(cowork-vm-shutdown): VM service not running.
20:39:27 [error] Session initialization failed: VM service not running.
20:39:49 [error] Session initialization failed: VM service not running.
20:41:10 [error] Sentry caught (x5)

Fix: manually starting the service

PS> Get-Service CoworkVMService
# Status: Stopped, StartType: Automatic

PS> Start-Service CoworkVMService
# Cowork immediately works

Totals since Apr 4

Steps to Reproduce

This is intermittent and the trigger for the service stopping is unknown. Pattern observed:

1. Use Cowork normally (service running)
2. At some point the CoworkVMService stops silently (no log entry for why)
3. Next Cowork interaction triggers crash-restart loop
4. App never recovers on its own

Expected Behavior

1. Auto-restart the service: If CoworkVMService is stopped when Cowork is needed, attempt Start-Service before giving up
2. Don't crash-loop: After detecting "VM service not running" once, show an actionable error ("Cowork service stopped. Click to restart.") instead of crash-restarting the entire app
3. Log why the service stopped: Currently there is no log entry indicating why the service dies — only downstream failures

Workaround

Added to the existing scheduled task (ClearCoworkVMs) that runs hourly:

# Also restart the service if it's stopped
$svc = Get-Service -Name 'CoworkVMService' -ErrorAction SilentlyContinue
if ($svc -and $svc.Status -ne 'Running') {
    Start-Service -Name 'CoworkVMService' -ErrorAction SilentlyContinue
}

This mitigates but doesn't eliminate the issue (can still hit it between hourly checks).

Environment

Log Files

• %APPDATA%\Claude\logs\main.log — crash cycles, shutdown failures, session init failures
• %APPDATA%\Claude\logs\main1.log — rotated log with Apr 8–11 history