claude-code - 💡(How to fix) Fix [BUG] Cowork fails with uv_spawn EUNKNOWN (errno -134) — VM boots successfully but Bun cannot spawn child processes [2 comments, 2 participants]

Preflight Checklist

• I have searched existing issues and this hasn't been reported yet
• This is a single bug report (please file separate reports for different bugs)
• I am using the latest version of Claude Code

What's Wrong?

Cowork VM boots successfully — network connects, API is reachable, SDK installs — but every session immediately crashes with exit code 1 because the Bun runtime (v1.3.13) inside the VM cannot spawn any child process. The underlying error is: EUNKNOWN: unknown error, uv_spawn syscall: "uv_spawn", errno: -134, code: "EUNKNOWN"

This happens on every attempt, regardless of whether a project/skill is selected or not. Even a blank Cowork session with no project triggers the same crash. Cowork was previously working on this machine. The issue started appearing after a Claude Desktop auto-update (exact version unknown). This is NOT the typical ENOENT spawn error seen in other issues. errno -134 is an unknown/unhandled error code from libuv inside Bun, suggesting something specific to the VM's execution environment is preventing process creation entirely.

What Should Happen?

After VM boots successfully (which it does — all steps complete in ~10 seconds, network CONNECTED, API REACHABLE, SDK installed), Cowork should be able to spawn child processes and start a working session. The user should be able to interact with Cowork normally.

Error Messages/Logs

2026-04-22 11:43:18 [info] [VM:start] Beginning startup, bundlePath=C:\Users\emre.demir\AppData\Roaming\Claude\vm_bundles\claudevm.bundle
2026-04-22 11:43:18 [info] [VM:start] Bundle version: 5680b11bcdab215cccf07e0c0bd1bd9213b0c25d
2026-04-22 11:43:18 [info] [VM:start] VM instance ID: 2a89e062-d6d5-44c3-802d-fe297384b36f
2026-04-22 11:43:18 [info] [VM:steps] download_and_sdk_prepare completed (18ms)
2026-04-22 11:43:18 [info] [VM:steps] load_swift_api completed (0ms)
2026-04-22 11:43:18 [info] [VM:start] smol-bin.x64.vhdx copied successfully
2026-04-22 11:43:18 [info] [VM:start] Windows VM service configured
2026-04-22 11:43:18 [info] [VM:steps] create_network completed (1ms)
2026-04-22 11:43:18 [info] [VM:steps] create_vm_config completed (257ms)
2026-04-22 11:43:19 [info] [VM:steps] vm_boot completed (226ms)
2026-04-22 11:43:19 [info] [VM:steps] add_plan9_shares completed (13ms)
2026-04-22 11:43:26 [info] [VM] Network status: CONNECTED
2026-04-22 11:43:26 [info] [VM:steps] sdk_install completed (1384ms)
2026-04-22 11:43:28 [info] [VM:start] Startup complete, total time: 10038ms
2026-04-22 11:43:28 [info] [VM] API reachability: REACHABLE

Main app log (main.log) — CRASH HERE ❌

2026-04-22 11:50:21 [warn] [PluginScan] Skipping legacy command "brand-voice:discover-brand" — name collides with skills/ entry

2026-04-22 11:50:21 [warn] [HostLoop] cli.js stderr:
674 |     var file = this.spawnfile = options.file, spawnargs;
675 |     if (options.args === @undefined)
676 |       spawnargs = this.spawnargs = [];
677 |     else
678 |       validateArray(options.args, "options.args"), spawnargs = this.spawnargs = options.args;
679 |       if (this.#handle = Bun.spawn({
                                          ^
EUNKNOWN: unknown error, uv_spawn
 syscall: "uv_spawn",
   errno: -134,
    code: "EUNKNOWN"

      at spawn (node:child_process:679:35)
      at spawn (node:child_process:14:39)
      at execFile (node:child_process:59:20)
      at <anonymous> (B:/~BUN/root/src/entrypoints/cli.js:187:1598)

Bun v1.3.13 (Windows x64 baseline)

2026-04-22 11:50:21 [error] Sentry caught: { type: 'Error', value: 'write EOF' }

2026-04-22 11:50:21 [error] Session query error: Claude Code process exited with code 1

2026-04-22 11:50:21 [info] [CycleHealth] Unhealthy cycle: {
  cycle_health: 'unhealthy',
  had_first_response: false,
  seconds_to_outcome: 1,
  unhealthy_reason: 'system_error',
  error_message: 'Claude Code process exited with code 1',
  error_category: 'process_crashed',
}

Steps to Reproduce

1. Open Claude Desktop (v1.3883.0) on Windows 11 Pro
2. Click on the Cowork tab
3. Wait for VM to boot (it boots successfully — ~10 seconds)
4. Try to start any Cowork session — with or without selecting a project
5. Session immediately crashes with "Claude Code process exited with code 1"

No specific files, code, or project selection is needed to reproduce. The crash occurs on every attempt, even with a completely blank session and no skills/projects configured.

Claude Model

Opus

Is this a regression?

Yes, this worked in a previous version

Last Working Version

Exact version unknown — Cowork was working on this same machine before a Claude Desktop auto-update around mid-April 2026. Current broken version is 1.3883.0.

Claude Code Version

Cannot run CLI commands (corporate managed machine, no terminal access). From logs: Claude Desktop 1.3883.0, bundled Claude Code SDK v2.1.111, Bun v1.3.13

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

PowerShell

Additional Information

Important Context This is a Cowork issue, not a CLI issue. I'm filing here because this appears to be the shared issue tracker. I have no terminal/CLI access — this is a corporate managed Windows machine.

Troubleshooting Attempted (all failed) Full clean reinstall from claude.com/download | ❌ Same error Deleted vm_bundles, claude-code-vm, local-agent-mode-sessions | ❌ VM rebuilt, same crash Deleted VHDX disk images | ❌ Recreated, same crash Restarted CoworkVMService manually | ❌ Service runs fine, crash persists Added Windows Defender exclusions for Claude paths | ❌ No change Checked Controlled Folder Access (disabled) | ❌ Not the cause Checked for ANTHROPIC_API_KEY env variable (none set) | ❌ Not the cause Checked corporate endpoint security software | ❌ No blocks detected Checked AppLocker/WDAC policies | ❌ No relevant policies Hyper-V NAT reset (removed and recreated) | ❌ No change VPN completely closed | ❌ No change Tested with and without a project selected | ❌ Same error in both cases Multiple PC reboots | ❌ No change

Key Technical Detail The errno: -134 in uv_spawn is not the typical ENOENT (-2) or EACCES (-13) seen in other issues. This is an undocumented/unhandled libuv error code, suggesting a Bun-specific or VM-specific regression in process spawning on Windows Hyper-V. Environment Details

Windows 11 Pro CoworkVMService: Running VM boots fully, network CONNECTED, API REACHABLE Corporate managed machine (no admin/terminal access for end user) Hyper-V and Virtual Machine Platform enabled No VPN interference

Related Issues

#49514 — Similar "exit code 1 persists after clean reinstall" on Windows 11 #49328 — "exit code 1 happening with cowork on Desktop" #49435 — "Cowork VM service fails to start after update" #39270 — "process exited with code 1" started during March 2026 outage