[BUG] Cowork: "Failed to create bridge sockets after 5 attempts" on every spawn after update to 1.3036.0 (cf521c)

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?

Every new Cowork task in Claude Desktop crashes within ~1 second with "Something went wrong — Claude Code process exited with code 1." The Linux VM spawns successfully, then the in-guest process fails to create its bridge sockets (5 retry attempts) and exits with code 1. Reproduces 100% of the time on every new task. Existing tasks also fail to resume. Claude Desktop Cowork is fully unusable on this build. Claude Chat and the Claude Code CLI from the same machine on the same auth are unaffected.

What Should Happen?

A new Cowork task should spawn the VM, initialize the in-guest bridge sockets, and become interactive within a few seconds, not crash with exit code 1 before the first user turn.

Error Messages/Logs

From ~/Library/Logs/Claude/cowork_vm_node.log — identical pattern repeats for every spawn attempt over 35+ minutes:                                                                                                
                                                                                                                     
  2026-04-16 11:38:41 [info] [Spawn:vm] id=6caac35f-... Spawn succeeded in 93ms                                                                                                                                      
  2026-04-16 11:38:41 [info] [Process:6caac35f-...] Spawn confirmed, flushing 2 buffered stdin chunks                                                                                                                
  2026-04-16 11:38:42 [warn] [vm-stderr 6caac35f] Error: Failed to create bridge sockets after 5 attempts                                                                                                            
  2026-04-16 11:38:42 [info] [Process:6caac35f-...] Exited, code=1, signal=null, duration=1175ms, oom=false                                                                                                          
  2026-04-16 11:38:42 [info] [Process:6caac35f-...] Cleaned up, remaining active=0                                                                                                                                   
                                                                                                                                                                                                                     
  Key signals:                                                                                                                                                                                                       
  - "Spawn succeeded" — the VM itself boots and the host-side process is created                                                                                                                                     
  - "Failed to create bridge sockets after 5 attempts" — the in-guest bridge setup retries and gives up                                                                                                              
  - duration ~1s, oom=false — deterministic early failure, not a memory issue

Steps to Reproduce

1. Update Claude Desktop to 1.3036.0 (cf521c), built 2026-04-16T16:17:33Z
2. Launch Claude Desktop, sign in (OAuth / Claude Team), and navigate to Cowork
3. Start a new Cowork task with any prompt (e.g. "are you functioning properly?")
4. Within ~1 second, the error appears: "Something went wrong — Claude Code process exited with code 1"

Claude Model

Opus

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

2.1.111 (Claude Code)

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Other

Additional Information

Claude Desktop version: 1.3036.0 (cf521c), built 2026-04-16T16:17:33Z
Hardware: Apple M1 Pro, arm64
Auth: OAuth via Claude Team — no ANTHROPIC_API_KEY in env

Scope: The Claude Code CLI (2.1.111) from the same machine, same ~/.claude/ settings, same OAuth session, works fine. Isolated to Claude Desktop's Cowork VM in this specific build.

Troubleshooting already attempted (no change):

• /doctor — reports clean
• which -a claude — single clean install at ~/.local/bin/claude → ~/.local/share/claude/versions/2.1.111
• Cleared stale /tmp/claude-501/* session directories
• Full Cmd-Q and relaunch
• Verified TMPDIR is valid and writable
• Confirmed no ANTHROPIC_API_KEY in environment
• status.claude.com — all systems operational

Related reports (cluster filed within ~20 min of the 1.3036.0 rollout, same error signature):

• macOS: #49328, #49341, #49346, #49358, #49360, #49366, #49367
• Windows: #49362, #49365

Pattern strongly suggests a build-wide regression, not an environment-specific issue. Filing this as a separate report because it includes the specific vm-stderr log signature and reproduces on Tahoe 26.4 / M1
Pro — happy to close as duplicate of #49328 if more useful consolidated there.

Logs available: Full cowork_vm_node.log, cowork_vm_swift.log, and coworkd.log available on request if there's an upload path.