[BUG] Cowork VM starts but API is UNREACHABLE on Windows 11

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?

Details:

Windows 11, Claude Desktop v1.1.2685
Hyper-V enabled, CoworkVMService running
VM downloads, boots, and gets CONNECTED status
API reachability always goes PROBABLY_UNREACHABLE → UNREACHABLE
Host can resolve and reach api.anthropic.com and a-cdn.anthropic.com fine

What Should Happen?

Failed to start Claude's workspace
Can't reach the Claude API from Claude's workspace.
Restarting Claude or reconnecting to your network sometimes resolves this.

Error Messages/Logs

2026-02-12 17:14:49 [info] Downloading vmlinuz...
2026-02-12 17:14:49 [info] vmlinuz.zst checksum validated
2026-02-12 17:14:50 [info] [Bundle:status] initrd missing
2026-02-12 17:14:50 [info] [warm] No warm file found for initrd at current version
2026-02-12 17:14:50 [info] initrd not found, downloading...
2026-02-12 17:14:50 [info] Downloading initrd...
2026-02-12 17:15:05 [info] initrd.zst checksum validated
2026-02-12 17:15:05 [info] All files ready in C:\Users\mcass\AppData\Roaming\Claude\vm_bundles\claudevm.bundle
2026-02-12 17:15:05 [info] [VM:steps] download_and_sdk_prepare completed (219534ms)
2026-02-12 17:15:05 [info] [VM:steps] load_swift_api started
2026-02-12 17:15:05 [info] [VM:steps] load_swift_api completed (0ms)
2026-02-12 17:15:05 [info] [Callbacks] Initializing VM event callbacks...
2026-02-12 17:15:05 [info] [Callbacks] VM event callbacks initialized
2026-02-12 17:15:05 [info] [VM:start] Copying smol-bin.vhdx to bundle: C:\Program Files\WindowsApps\Claude_1.1.2685.0_x64__pzs8sxrjxfjjc\app\resources\smol-bin.vhdx -> C:\Users\mcass\AppData\Roaming\Claude\vm_bundles\claudevm.bundle\smol-bin.vhdx
2026-02-12 17:15:05 [info] [VM:start] smol-bin.vhdx copied successfully
2026-02-12 17:15:05 [info] [VM:start] Configuring Windows VM service...
2026-02-12 17:15:06 [info] [VM:start] Windows VM service configured
2026-02-12 17:15:16 [info] [VM] Network status: NOT_CONNECTED
2026-02-12 17:15:16 [info] [VM:network] Not connected, will show error in 30000ms if not resolved
2026-02-12 17:15:16 [info] [VM] Network status: CONNECTED
2026-02-12 17:15:17 [info] [VM:steps] sdk_install started
2026-02-12 17:15:17 [info] [VM:start] Installing SDK: subpath=AppData/Roaming/Claude/claude-code-vm, version=2.1.34
2026-02-12 17:15:18 [info] [VM:steps] sdk_install completed (1773ms)
2026-02-12 17:15:18 [info] [VM:start] Startup complete, total time: 233054ms
2026-02-12 17:15:18 [info] [Keepalive] Starting (interval=2000ms)
2026-02-12 17:15:21 [info] [VM] API reachability: PROBABLY_UNREACHABLE
2026-02-12 17:15:45 [info] [VM] API reachability: UNREACHABLE
2026-02-12 17:15:45 [warn] [VM:network] API is unreachable
2026-02-12 17:21:01 [info] [VM:shutdown] App quit, stopping VM (instance: 49f9de98-938e-4f7c-8df9-d05fbb8cdb93)...
2026-02-12 17:21:01 [info] [Keepalive] Stopping
2026-02-12 17:21:02 [info] [VM:shutdown] Completed in 480ms
2026-02-12 17:21:11 [info] [VM] Loading vmClient (TypeScript) module...
2026-02-12 17:21:11 [info] [VM] Module loaded successfully
2026-02-12 17:21:13 [info] [VM:start] Beginning startup, bundlePath=C:\Users\mcass\AppData\Roaming\Claude\vm_bundles\claudevm.bundle
2026-02-12 17:21:13 [info] [VM:start] Bundle version: fc96f5bc6a7aeee6e651437eba3c76d0fe9a16f3
2026-02-12 17:21:13 [info] [VM:start] VM instance ID: f53501bd-3b21-4949-8a25-4842afe94104
2026-02-12 17:21:13 [info] [VM:steps] download_and_sdk_prepare started
2026-02-12 17:21:13 [info] [VM:steps] download_and_sdk_prepare completed (16ms)
2026-02-12 17:21:13 [info] [VM:steps] load_swift_api started
2026-02-12 17:21:13 [info] [VM:steps] load_swift_api completed (0ms)
2026-02-12 17:21:13 [info] [Callbacks] Initializing VM event callbacks...
2026-02-12 17:21:13 [info] [Callbacks] VM event callbacks initialized
2026-02-12 17:21:13 [info] [VM:start] Copying smol-bin.vhdx to bundle: C:\Program Files\WindowsApps\Claude_1.1.2685.0_x64__pzs8sxrjxfjjc\app\resources\smol-bin.vhdx -> C:\Users\mcass\AppData\Roaming\Claude\vm_bundles\claudevm.bundle\smol-bin.vhdx
2026-02-12 17:21:13 [info] [VM:start] smol-bin.vhdx copied successfully
2026-02-12 17:21:13 [info] [VM:start] Configuring Windows VM service...
2026-02-12 17:21:14 [info] [VM:start] Windows VM service configured
2026-02-12 17:21:20 [info] [VM] Network status: NOT_CONNECTED
2026-02-12 17:21:20 [info] [VM:network] Not connected, will show error in 30000ms if not resolved
2026-02-12 17:21:20 [info] [VM:steps] sdk_install started
2026-02-12 17:21:20 [info] [VM:start] Installing SDK: subpath=AppData/Roaming/Claude/claude-code-vm, version=2.1.34
2026-02-12 17:21:20 [info] [VM] Network status: CONNECTED
2026-02-12 17:21:22 [info] [VM:steps] sdk_install completed (2673ms)
2026-02-12 17:21:22 [info] [VM:start] Startup complete, total time: 9170ms
2026-02-12 17:21:22 [info] [Keepalive] Starting (interval=2000ms)
2026-02-12 17:21:25 [info] [VM] API reachability: PROBABLY_UNREACHABLE
2026-02-12 17:21:49 [info] [VM] API reachability: UNREACHABLE
2026-02-12 17:21:49 [warn] [VM:network] API is unreachable

Steps to Reproduce

i don't know what steps to take. cowork is not connecting to the api. I think we've exhausted what we can reasonably fix on our end. The pattern is 100% consistent across every attempt: the VM boots, gets network connectivity to the host, but cannot reach the Anthropic API from inside the VM. This is the CoworkVMService's internal NAT not routing traffic to the internet properly — and that's something only Anthropic can fix.

Claude Model

Opus

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

Version 1.1.2685 (f39a62)

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

Windows Terminal

Additional Information

_No response_