[BUG] Cowork fails - VM cannot reach Claude API, Default Switch is Internal type with no external uplink (Windows 11 Pro, v1.3883.0)

Resolved 💬 2 comments Opened Apr 22, 2026 by wenhoreda Closed May 29, 2026

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 consistently fails to start with "Can't reach the Claude API from Claude's workspace." Regular Chat mode works perfectly. A yellow validation_failed badge appears briefly in the Cowork tab.

What Should Happen?

Cowork should either create an External virtual switch linked to the active network adapter, or configure NAT correctly so the VM can reach api.anthropic.com.

Error Messages/Logs

Can't reach the Claude API from Claude's workspace

Steps to Reproduce

Start Cowork - it fails

Claude Model

Not sure / Multiple models

Is this a regression?

No, this never worked

Last Working Version

_No response_

Claude Code Version

1.3883.0 (93ff6c), 2026-04-21

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

PowerShell

Additional Information

Troubleshooting performed
Hyper-V & services — all OK:

Hyper-V enabled (Microsoft-Hyper-V-All → State: Enabled)
CoworkVMService: Running
vmcompute: Running
vmms: Running
Virtualization enabled in BIOS (HyperVRequirementVirtualizationFirmwareEnabled: True)

Network — host reaches API fine:
Test-NetConnection -ComputerName api.anthropic.com -Port 443
TcpTestSucceeded: True (160.79.104.10)
Network adapters (sorted by metric):
Ethernet IPv4 metric 4 Disconnected
Wi-Fi IPv4 metric 10 Connected
vEthernet (Default Switch) IPv4 metric 5000 Connected
vEthernet (Default Switch) IPv6 metric 5000 Connected
Root cause identified:
Get-VMSwitch
Name SwitchType NetAdapterInterfaceDescription
Default Switch Internal
The Hyper-V Default Switch is type Internal — it has no external adapter assigned as uplink. The Cowork VM boots and gets network status CONNECTED internally, but cannot route traffic to api.anthropic.com because the switch has no path to the physical network.
Steps already tried without success:

Updated Claude Desktop to latest version (1.3883.0)
Enabled Hyper-V + reboot
Cleared cache via Help → Troubleshooting → Clear cache and restart
Deleted vm_bundles and claude-code-vm from %APPDATA%\Claude + restarted HNS and vmcompute
Adjusted network adapter metrics (Wi-Fi to 10, Ethernet to 4)
Full PC reboot after each attempt

Expected behavior
Cowork should either create an External virtual switch linked to the active network adapter, or configure NAT correctly so the VM can reach api.anthropic.com.
Actual behavior
Default Switch remains Internal with no uplink. VM boots but API is unreachable. Error: "Can't reach the Claude API from Claude's workspace."
Question for the team
Should Cowork automatically create an External switch or configure NAT on setup? Is there a manual workaround that doesn't require permanently bridging the physical Wi-Fi adapter?

View original on GitHub ↗

This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗