[BUG] 1.1617.0 (8d6345): New Cowork sessions crash with "exited with code 1" - /doctor and /bug non-functional

Reset App Data.

Did nothing other than nuking my progress, all my sessions. For nothing.

The key points for the issue:

• Build 1.1617.0 / Claude Code 2.1.92 crashes with "exited with code 1" on new Cowork sessions when Git for Windows is not installed
• The actual error is buried in cowork_vm_node.log: "Claude Code was unable to find CLAUDE_CODE_GIT_BASH_PATH path bash.exe"
• No user-facing error message — just a crash. Previous builds didn't require Git for session creation
• Existing sessions resumed fine, masking the root cause
• /doctor and /bug also crash, so users can't even self-diagnose

### - Fix: install Git for Windows and add to PATH

What should have happened: a dialog saying "Git is required. Install it from git-scm.com" — not a silent crash that sent you down a two-hour rabbit hole of cache clearing, reset, backups, and rollback attempts.

Fix: install Git for Windows and add to PATH

Claude Code process exited with code 1
You can restart the conversation from an earlier message.

Additional report: Mac Sequoia, same build

Version: Claude 1.1617.0 (8d6345) 2026-04-09T16:10:15.000Z
OS: macOS Sequoia
Behavior: Opposite of the Windows report above -- new sessions accept input and work normally, but existing/prior sessions have a completely frozen prompt input field with no error indicator
Restarting Claude does not resolve it

Title: Cowork new sessions crash with "exited with code 1" on Windows 10
Body:

Version: Claude 1.1617.0 (8d6345) 2026-04-09T16:10:15.000Z
OS: Windows 10 (English)
What's wrong:
New Cowork tasks fail immediately with "Claude Code process exited with code 1". The error appears after ~1 second. Existing/previous tasks also disappeared after clearing session data from %APPDATA%\Claude.
What I tried:

Restarted Claude Desktop
Restarted computer
Uninstalled and reinstalled Claude Desktop
Deleted Session Storage and local-agent-mode-sessions from %APPDATA%\Claude
Confirmed no ANTHROPIC_API_KEY environment variable set

None of the above resolved the issue.

I Am also suffering from this issue. installing git for windows did not help, neither did all the other ideas mentioned above

same issue here with CoWork: just throws out this error message starting this morning.

Tried a variety of ways to fix this that Claude recommended, but no luck: cleared config, upgraded Claude Code, fixed plugin errors, fixed the Claude-3p permission, tried a different model. Claude Code works from the terminal.

App version: 1.3036.0
OS: macOS, Apple Silicon (Darwin arm64)
Node: 24.14.0
Claude Code (Homebrew): 2.1.92
Bundled SDK: 2.1.111
Crash reproduces on Opus 4.7 and Sonnet
Crash is 1 second after session init, error_category: 'process_crashed', no stderr reaches the UI
Survives full config wipe (~/.claude, ~/.claude.json, Claude-3p permission fix)

Still broken on macOS, version 1.3036.0 (build 1.3036.0), SDK 2.1.111

Same symptom — every Cowork process spawn fails with Failed to create bridge sockets after 5 attempts and exits with code 1 within ~800–900ms. Affects both new sessions and resume of existing ones.

Environment:

• macOS 25.3 (Apple Silicon)
• Claude Desktop 1.3036.0
• Bundled SDK 2.1.111
• VM bundle version: 5680b11bcdab215cccf07e0c0bd1bd9213b0c25d

Log excerpt (~/Library/Logs/Claude/cowork_vm_node.log):

[Spawn:vm] id=67f94bb2 OAuth token approved with MITM proxy
[Spawn:vm] id=67f94bb2 Spawn succeeded in 75ms
[vm-stderr 67f94bb2] Error: Failed to create bridge sockets after 5 attempts
[Process:67f94bb2] Exited, code=1, signal=null, duration=904ms, oom=false

What I tried (none helped):

• Full quit + restart of Claude.app
• Deleted ~/Library/Application Support/Claude/vm_bundles (forced fresh rootfs.img download)
• Deleted ~/Library/Application Support/Claude/claude-code-vm (forced SDK reinstall)
• Confirmed TMPDIR is writable on host
• Confirmed ANTHROPIC_API_KEY is unset (no auth conflict)

VM boots cleanly, OAuth approved, network reachable — failure is deterministic at the bridge-socket creation step inside the guest.

Downgrading to an older Claude Desktop build (1.1.5749) avoids the crash but loses the Cowork Projects UI state.

Same issue for new and existing projects in Claude Cowork.

Cowork is broken for me now. none of my projects run

Same issue here.

I am seeing this issue for new members our team -- and they are on Mac M1-M4

This issue needs to be reopened

Workaround that worked for me

Since the bug isn't fixed yet in 1.3036.0, I rolled the desktop app back to an older build I had lying around (1.1.5749 from March). Steps:

# 1. Quit Claude
killall Claude

# 2. Backup current (broken) version
mv /Applications/Claude.app /Applications/Claude.app.backup-1.3036.0

# 3. Mount older DMG and copy
hdiutil attach ~/Downloads/Claude.dmg -nobrowse -quiet
cp -R /Volumes/Claude/Claude.app /Applications/
hdiutil detach /Volumes/Claude -quiet

# 4. Launch
open -a Claude

Trade-off: the older build doesn't render the Cowork Projects UI (it predates that feature), but Cowork sessions themselves work — no more Failed to create bridge sockets. Project files and IndexedDB metadata are preserved on disk, so once a fixed 1.30xx.x ships, upgrading should restore the Projects view.

Where the data lives, for anyone who needs to verify nothing is lost during the downgrade:

• Project files: ~/Documents/Claude/Projects/
• Project metadata: ~/Library/Application Support/Claude/IndexedDB/https_claude.ai_0.indexeddb.leveldb/
• Session data: ~/Library/Application Support/Claude/claude-code-sessions/

Hoping for a 1.30xx.x patch soon.

I'm not a Claude Code user but an avid Cowork user and am getting this same error. Tried all the fixes suggested above here with no luck. Question @NSSMedia if I try this older build, will I lose project data I've worked on today you think? As in chat history, progress, etc? I was mid Website build for sales dashboards and am hoping to recover as much as I can. Worth the fix either way but want to be prepared

I'm also experiencing the same right now. It's frustrating. It happened after I relaunched it cos of the update notification

Same issue here on Windows. Cowork desktop, Opus 4.6. After the recent update, new sessions crash immediately with "process exited with code 1". Additionally, resuming existing sessions also fails — error appears as soon as the agent starts responding. This breaks production workflows with autonomous CRON agents. Clear cache + restart didn't help. No rollback option available.

same @edd-virtualexec I'm distraught :'( four hour project potentially lost. never hitting the relaunch with update button again!

same @edd-virtualexec I'm distraught :'( four hour project potentially lost. never hitting the relaunch with update button again!

Oh my God. The thing is today is the day where I have lots of things to do with it. LOL perfect timing. do you guys know when this issue gets fixed usually? I did a couple of research and apparently, something like this happened last month too.

It's frustrating. But the CHAT works thought. its just that I built a lot already in my Cowork workspace. HAHAHA

Cowork aint taking over any jobs anytime soon lol

that part!

Samme issue here after last update, Mac OS Tahoe 26.3.1

+1，macOS 26.4.1 + Claude Desktop 1.3036.0, same problem

<img width="379" height="166" alt="Image" src="https://github.com/user-attachments/assets/ab31f2c8-72ae-4434-8206-482d1b46c2fd" />

Well, im not doing anything yet. There's an option to revert to the previous build but I'm too afraid to do anything for now. Afraid I'll lost some things doing that option.

Any expert tech persons here? is doing that safe? I was thinking to just wait for Anthropic to fix this since its a global user issue after the update anyway. LOL

Confirming on macOS 1.3036.0 (cf521c), SDK 2.1.111, VM bundle 5680b11bcdab215cccf07e0c0bd1bd9213b0c25d. Reproduces across Opus 4.7, Opus 4.7[1m], Opus 4.6, and Sonnet 4.6, on both new and resumed sessions. Same ~800-1000ms duration before exit 1.

Host-side VM is healthy in my logs - VM boots in ~90ms, network connects, API reachable, OAuth approved, spawn confirmed. Failure is consistently inside the guest:

[Spawn:vm] id=afa1c65b OAuth token approved with MITM proxy
[Spawn:vm] id=afa1c65b Spawn succeeded in ~100ms
[vm-stderr afa1c65b] Error: Failed to create bridge sockets after 5 attempts
[Process:afa1c65b] Exited, code=1, signal=null, duration=~900ms, oom=false

Tried: full quit + relaunch, sign out / sign in, Help -> Troubleshooting -> Clear Cache and Restart. No effect. Confirmed no ANTHROPIC_API_KEY in environment. Started immediately after auto-update to 1.3036.0 coinciding with Opus 4.7 rollout.

Follow-up: not caused by remote plugins

Ruled out remote plugins as the trigger. Renamed the session's rpm/ folder (~/Library/Application Support/Claude/local-agent-mode-sessions/<session>/<subsession>/rpm) to disable all 8 remote plugins, relaunched.

Log confirms plugins were unloaded:

[Spawn:config] Creating spawn function for process=zealous-compassionate-johnson, isResume=false, mounts=7 (outputs, .claude, .claude/skills, .cowork-lib, .cowork-perm-req, .cowork-perm-resp, uploads), allowedDomains=23

Down from mounts=18 previously.

Still crashes identically:

[vm-stderr] Error: Failed to create bridge sockets after 5 attempts
[Process] Exited, code=1, signal=null, duration=909ms, oom=false

Same duration (~900ms), same error. Issue is in baseline VM bridge-socket setup, not user-plugin surface area.

I am also seeing the same error on MacOS. Why is this issue closed? Can someone re-open this?

<img width="492" height="209" alt="Image" src="https://github.com/user-attachments/assets/e807e599-0d89-47e3-94c8-d4034f8f64c2" />

Live in all cowork attempts, up to date mac os on M5 MBA

Cowork appears to be broken for everyone running the latest build. I would think this would be treated with highest priority.

Same issue here. Since the latest update, every new Cowork session crashes immediately with "Claude Code process exited with code 1". Tried restarting and clearing cache — no luck.

OS: macOS
Version: latest

Getting code 1 pop up and also this when starting a new cowork env.

<img width="699" height="181" alt="Image" src="https://github.com/user-attachments/assets/337076bb-c20d-4d61-84eb-b6e95f110f32" />

Now I have to do actual work ☹️

<img width="999" height="258" alt="Image" src="https://github.com/user-attachments/assets/1e5ea24f-62a3-4761-bc4e-999a13c2d7aa" />

This is being tracked on the Claude Status page now.

Hoping this gets fixed and I've joind the club right after updating the latest version of desktop app on windows (updated around 3:20PM EST on April 16th, 2026). Can't do anything in cowork but chat is fine.

Same issue, Cowork update to latest version -> Existing and new chats.

<img width="531" height="269" alt="Image" src="https://github.com/user-attachments/assets/0d2248bf-fe6b-437d-8bee-45afe61e0a62" />

Can confirm this issue as well on Co-Work
<img width="1210" height="512" alt="Image" src="https://github.com/user-attachments/assets/30b4a6ef-849c-4cb6-8d03-badbf70e61f2" />

Claude 1.3036.0 (cf521c) 2026-04-16T16:17:33.000Z
ProductName: macOS
ProductVersion: 26.4
BuildVersion: 25E246

Same issue for me.

Same issue, Cowork update to latest version -> Existing and new chats.
<img width="356" height="161" alt="Image" src="https://github.com/user-attachments/assets/13b9fd32-b2fd-4fe8-a495-713cb7d629fe" />

ProductName: macOS M3
ProductVersion: 26.4.1（25E253）

same issue

had the same issue, this helped with Windows:

• installed Git from git-scm.com
• Created file C:\Users\YourUsername\.claude\settings.json

Content in the json file:
{
"env": {
"CLAUDE_CODE_GIT_BASH_PATH": "C:\\Program Files\\Git\\bin\\bash.exe"
}
}

Same issue, Cowork update to latest version -> Existing and new chats. <img alt="Image" width="356" height="161" src="https://private-user-images.githubusercontent.com/26702954/579489076-13b9fd32-b2fd-4fe8-a495-713cb7d629fe.png?jwt=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.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.l_FkS0R3BmayNmKM_bQuF-54SFjdC_vIduG15BejnNY"> ProductName: macOS M3 ProductVersion: 26.4.1（25E253）

This is exactly my issue as well.

Update from Claude status page

<img width="869" height="228" alt="Image" src="https://github.com/user-attachments/assets/b930dde5-f420-433e-9b72-4d38a24d823c" />

Fixed for me with the last update

This issue has been automatically locked since it was closed and has not had any activity for 7 days. If you're experiencing a similar issue, please file a new issue and reference this one if it's relevant.