[BUG] Cowork workspace bricked after Chrome automation task on Windows 11 Home, no recovery path

This automated invalid tag is BS

Found 3 possible duplicate issues:

1. https://github.com/anthropics/claude-code/issues/25419
2. https://github.com/anthropics/claude-code/issues/25132
3. https://github.com/anthropics/claude-code/issues/25206

This issue will be automatically closed as a duplicate in 3 days.

• If your issue is a duplicate, please close it and 👍 the existing issue instead
• To prevent auto-closure, add a comment or 👎 this comment

🤖 Generated with Claude Code

Example 1 is with PRO not Home
Example 2 is the most similar but seems to be from first install. Mine was working fine but failed after a complex task, suggesting regular use could cause fatal error
Example 3 is also PRO

UPDATE: After updating to Windows Pro it seems to be fixed. Will update if this changes.

Had the exact same error:

sandbox-helper: host share not mounted at /mnt/.virtiofs-root/shared: not a mount point

Clearing cache and restarting did NOT fix it.

What fixed it: Enabling "Windows Hypervisor Platform" in Windows Features.

Steps:

1. Open "Turn Windows features on or off"
2. Check "Windows Hypervisor Platform" (I already had "Virtual Machine Platform" enabled, but not this one)
3. Restart PC
4. Open Claude Desktop → Cowork worked immediately

Hope this helps someone. Windows 11 Pro here as well.

Update: I'm on Windows 11 Pro and had the same recurring issue.

Enabling only "Windows Hypervisor Platform" fixed it temporarily, but the error came back after a while.

Permanent fix (Windows Pro users): Enable full Hyper-V in "Turn Windows features on or off".

Make sure all three are enabled:

• ✅ Hyper-V (both sub-options: Management Tools + Hyper-V Platform)
• ✅ Virtual Machine Platform
• ✅ Windows Hypervisor Platform

Restart PC. Cowork has been stable since.

For Windows Home users who don't have access to full Hyper-V, the issue may still persist — this might explain why the VM keeps crashing on Home editions.

Same issue resolved — root cause: Tailscale VPN client

Environment:

• OS: Windows 11 Business (Pro-based), Build 22H2
• Hyper-V: Fully enabled (all components)
• Virtual Machine Platform: Enabled
• Windows Hypervisor Platform: Enabled
• Subscription: Max plan
• VPN: Tailscale client active

Symptoms experienced (in sequence):

1. Initial error — identical to OP:

CLI output was not valid JSON. This may indicate an error during startup.
Output: sandbox-helper: host share not mounted at /mnt/.virtiofs-root/shared: not a mount point

1. After deleting VM bundle and reinstalling workspace, error changed to:

Failed to start Claude's workspace
failed to create network endpoint: HcnCreateEndpoint failed: hr=0x80070490, error=
{"Success":false,"Error":"Element not found.","ErrorCode":2147943568}

Root cause: Tailscale VPN client was running. Its virtual network adapter interferes with Hyper-V HNS (Host Networking Service) endpoint creation for the Cowork VM.

Resolution:

1. Fully exit Tailscale (not just disconnect — Exit from System Tray)
2. Reset HNS data:

Stop-Service hns -Force
Remove-Item "C:\ProgramData\Microsoft\Windows\HNS\HNS.data" -Force
Start-Service hns

1. Restart computer
2. Launch Claude Desktop → Cowork works normally

Note: This issue was diagnosed and resolved with the help of Claude (Chat, Opus 4.6) through a step-by-step troubleshooting session — from identifying the virtiofs error, through HNS reset, to isolating Tailscale as the root cause.

Suggestion to Anthropic: Consider adding VPN detection during Cowork startup with a clear warning message (e.g., "VPN client detected — this may interfere with Claude's workspace. Please disable your VPN and retry.").

Same issue on Windows 11 Pro — VM boots but API unreachable, VPN adapters interfere with HCN

Environment

• OS: Windows 11 Pro 10.0.22631 (x64)
• CPU: 12th Gen Intel Core i7-12700H
• RAM: 16 GB
• Claude Desktop: 1.1.3189.0 (x64)
• Claude Code SDK: 2.1.41
• Hyper-V: Fully enabled (all components including HypervisorPlatform)
• VPN software installed: ExpressVPN (TAP + TUN adapters), Fortinet VPN (NDIS 6.30 + SSL adapters) — these are required for work and cannot be permanently uninstalled

Problem

Cowork workspace fails with multiple errors. The VM eventually boots successfully but the Claude API remains unreachable from inside the VM.

Error progression

1. Initial: sandbox-helper: host share not mounted at /mnt/.virtiofs-root/shared: not a mount point — Fixed by enabling HypervisorPlatform
2. Then: HcnCreateEndpoint failed with HRESULT 0x800706bf — Fixed after netsh winsock reset + netsh int ip reset + restart
3. Current: VM boots, network shows CONNECTED, but API goes PROBABLY_UNREACHABLE → UNREACHABLE within seconds

Key log evidence (cowork_vm_node.log)

19:32:24 [info] [VM:start] Windows VM service configured
19:32:34 [info] [VM] Network status: NOT_CONNECTED
19:32:34 [info] [VM:start] Installing SDK...
19:32:34 [info] [VM] Network status: CONNECTED
19:32:38 [info] [VM:start] Startup complete, total time: 124579ms
19:32:39 [info] [Spawn:vm] OAuth token approved with MITM proxy
19:32:39 [info] [VM] API reachability: PROBABLY_UNREACHABLE
19:33:07 [info] [VM] API reachability: UNREACHABLE
19:33:07 [warn] [VM:network] API is unreachable

Network diagnostics bug

The diagnostic code runs macOS commands on Windows:

[info] === Comprehensive macOS Diagnostics ===
[error] Failed to run /sbin/ifconfig: spawn /sbin/ifconfig ENOENT
[error] Failed to run /usr/sbin/scutil: spawn /usr/sbin/scutil ENOENT
[error] Failed to run /usr/bin/pgrep: spawn /usr/bin/pgrep ENOENT
[error] Failed to run /usr/sbin/kextstat: spawn /usr/sbin/kextstat ENOENT

This means zero useful network diagnostics are gathered on Windows.

Exhaustive troubleshooting attempted (all failed)

1. Enabled HypervisorPlatform (was Disabled)
2. Reset HNS data multiple times
3. Removed/recreated HNS networks
4. Created NetNat manually for 172.16.0.0/24
5. Removed Default Switch ICS to eliminate NAT conflict
6. netsh winsock reset + netsh int ip reset + restart
7. Disabled all VPN adapters (ExpressVPN TAP/TUN, Fortinet NDIS/SSL)
8. Removed vm_bundles completely (multiple times)
9. Full uninstall + clean reinstall (x64)
10. Tried ARM64 version, then back to x64
11. Help > Troubleshooting > Reinstall Workspace
12. Added firewall rules for vmcompute.exe
13. Restarted services (hns, vmcompute, vmms, WinNat) in every combination
14. Multiple full computer restarts

Critical issue: VPN adapter drivers

The VPN adapters (ExpressVPN TAP/TUN, Fortinet NDIS) interfere with HCN endpoint creation even when disabled. The kernel-level drivers are still loaded. However, these VPNs are required for work and cannot be uninstalled. Cowork needs to coexist with corporate VPN installations.

Suggestions to Anthropic

1. Fix network diagnostics on Windows — use ipconfig, netsh, Get-NetAdapter instead of macOS equivalents (ifconfig, scutil, kextstat)
2. Add VPN adapter detection at startup with a clear warning
3. Handle NetNat/ICS conflicts — New-NetNat fails with "duplicate name" error when ICS (Default Switch) is active
4. Improve Windows NAT routing — the VM creates cowork-vm-nat network but traffic doesn't route to internet

Same problem and indeed it was fixed with turning off TailScale

Closing for now — this doesn't appear to be about Claude Code. Please open a new issue if this is still relevant.

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.