[BUG]

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?

API reachability: UNREACHABLE after OAuth token approved with MITM proxy,

• Windows 11, Hyper-V/WSL2
• Claude v1.1.3363.0 (Microsoft Store)
• Wi-Fi connection (Realtek RTL8852BE)
• No VPN or third-party antivirus installed
• VM network status shows CONNECTED, SDK installs successfully
• Only the MITM proxy → API route fails
• Cleaning up stale cowork-vm-nat NAT/adapter did not resolve
• WSL2 can reach api.anthropic.com with no issues

What Should Happen?

After the VM starts and network status shows CONNECTED, the API reachability check should
return REACHABLE, allowing the Cowork workspace to connect to the Claude API and function
normally. Instead, it consistently goes PROBABLY_UNREACHABL

Error Messages/Logs

2026-02-18 09:07:38 [info] [VM:start] Beginning startup,
  bundlePath=C:\Users\beneh\AppData\Roaming\Claude\vm_bundles\claudevm.bundle
  2026-02-18 09:07:38 [info] [VM:start] Bundle version: 0ae58222a1f34cf74c74ce47cc717a82a4eb0f25
  2026-02-18 09:07:39 [info] [VM:start] Windows VM service configured
  2026-02-18 09:07:45 [info] [VM] Network status: NOT_CONNECTED
  2026-02-18 09:07:45 [info] [VM:network] Not connected, will show error in 30000ms if not
  resolved
  2026-02-18 09:07:45 [info] [VM:steps] sdk_install started
  2026-02-18 09:07:45 [info] [VM] Network status: CONNECTED
  2026-02-18 09:07:47 [info] [VM:steps] sdk_install completed (2109ms)
  2026-02-18 09:07:47 [info] [VM:start] Startup complete, total time: 8904ms
  2026-02-18 09:07:47 [info] [Keepalive] Starting (interval=2000ms)
  2026-02-18 09:07:50 [info] [VM] API reachability: PROBABLY_UNREACHABLE
  2026-02-18 09:08:05 [info] [Spawn:vm] id=0c0b07cc OAuth token approved with MITM proxy
  2026-02-18 09:08:14 [info] [VM] API reachability: UNREACHABLE
  2026-02-18 09:08:14 [warn] [VM:network] API is unreachable

  Pattern repeats on every startup attempt. VM has network (SDK installs fine) but API is always
   unreachable through the MITM proxy.

Steps to Reproduce

1. Open Claude desktop app (v1.1.3363.0, Microsoft Store) on Windows 11
2. Navigate to Cowork tab
3. Click to start a workspace / select a task
4. VM starts, network status transitions to CONNECTED
5. SDK installs successfully (v2.1.41)
6. OAuth token approved with MITM proxy
7. API reachability goes PROBABLY_UNREACHABLE → UNREACHABLE
8. Error displayed: "Can't reach the Claude API from Claude's workspace"
9. Issue reproduces 100% of the time across multiple restarts and days
10. Confirmed host machine can reach api.anthropic.com from both Windows and WSL2 (curl

succeeds)

Claude Model

None

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

Version 1.1.3363 (ee4247)

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

WSL (Windows Subsystem for Linux)

Additional Information

_No response_