[BUG] Cowork crashes on M4 Mac - seccomp killed error
Status Closed — not planned
Maintainer reply None cached
Workaround ✓ Mentioned in thread ↓
Activity 12 comments · opened Jan 18, 2026 · closed Apr 6, 2026
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?
Cowork fails to start on M4 Mac. It's using the Linux seccomp binary instead of macOS sandbox-exec.
What Should Happen?
Cowork should start normally using macOS sandbox-exec
Error Messages/Logs
CLI output was not valid JSON. This may indicate an error during startup. Output: /usr/bin/bash: line 4: 5 Killed /usr/local/lib/node_modules_global/lib/node_modules/@anthropic-ai/sandbox-runtime/vendor/seccomp/arm64/apply-seccompSteps to Reproduce
- Open Cowork on M4 Mac
- Error appears immediately on startup
Claude Model
Opus
Is this a regression?
I don't know
Last Working Version
_No response_
Claude Code Version
N/A - using Cowork (browser), not Claude Code CLI
Platform
Anthropic API
Operating System
macOS
Terminal/Shell
Terminal.app (macOS)
Additional Information
M4 Mac, sandbox-runtime updated to latest version via npm
12 Comments
Was there any update on this?
nothing, same issue for you?
Same problem on my side 👍
also have the same error
CLI output was not valid JSON. This may indicate an error during startup. Output: /usr/bin/bash: line 4: 5 Killed /usr/local/lib/node_modules_global/lib/node_modules/\@anthropic-ai/sandbox-runtime/vendor/seccomp/arm64/apply-seccomp /usr/local/lib/node_modules_g...
Also have the same error when trying to fire up cowork
CLI output was not valid JSON. This may indicate an error during startup. Output: /usr/bin/bash: line 4: 5 Killed /usr/local/lib/node_modules_global/lib/node_modules/\@anthropic-ai/sandbox-runtime/vendor/seccomp/arm64/apply-seccomp /usr/local/lib/node_modules_g...
We also have a similar error while working in a cowork session (M4) MBP, with no resolution until failure and having to move into another cowork chat (eventually happening again): "CLI output was not valid JSON. This may indicate an error during startup. Output: /usr/bin/bash: line 4: 5 Killed /usr/local/lib/node_modules_global/lib/node_modules/\@anthropic-ai/sandbox-runtime/vendor/seccomp/arm64/apply-seccomp /usr/local/lib/node_modules_g..."
Same problem here as well:
Environment:
App: Cowork
Package: @anthropic-ai/claude-code
Versions tested: 2.1.62 and 2.1.63 — both affected
OS: macOS (Apple Silicon) M4 Max
App: Cowork
Package: @anthropic-ai/claude-code
Versions tested: 2.1.62 and 2.1.63 — both affected
OS: macOS (Apple Silicon)
RAM: 128GB
SIP: enabled
Encounters the below error each time a new prompt is given:
Something went wrong
CLI output was not valid JSON. This may indicate an error during startup. Output: /usr/bin/bash: line 4: 5 Killed /usr/local/lib/node_modules_global/lib/node_modules/\@anthropic-ai/sandbox-runtime/vendor/seccomp/arm64/apply-seccomp /usr/local/lib/node_modules_g...
Still hitting this on the latest Claude Desktop build (March 2026).
Error:
What I've tried:
~/Library/Application Support/Claude/vm_bundles/claudevm.bundle/to force a fresh VM bundle rebuild → same error{"sandbox": {"enabled": false}}to~/.claude/settings.json→ setting appears to be ignored, same error persists/usr/local/lib/node_modules_global/...) does not exist on the Mac host — it resolves inside the VMConclusion: The error is happening inside the Ubuntu VM after it boots, not on the host. Since a fresh
rootfs.imgreproduces the issue immediately, the bug appears to be baked into the base VM image. There is no user-serviceable workaround available.Platform: macOS, Apple Silicon (ARM64)
Can this be prioritized? The issue is marked stale but is clearly still affecting users on current builds.
any fix available so far??
I'm also experiencing this issue. Adding my diagnostic details:
Environment:
Error:
What I tried (nothing worked):
rm -rf ~/Library/Caches/com.anthropic.claudefordesktoprm -rf ~/Library/Application\ Support/Claude/vm_bundles(VM recreated but same crash)The
apply-seccomppath only exists inside the VM'srootfs.img, not on the host. The rootfs origin hash isfb30784dadb34104626c8cf6d8f90dd47cd393cc. Signal 9 (SIGKILL) on every startup attempt.Closing for now — inactive for too long. Please open a new issue if this is still relevant.
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.