Interactive mode crashes with "zsh: suspended (signal)" on macOS 26.2

Found 3 possible duplicate issues:

1. https://github.com/anthropics/claude-code/issues/21567
2. https://github.com/anthropics/claude-code/issues/21806
3. https://github.com/anthropics/claude-code/issues/22196

This issue will be automatically closed as a duplicate in 3 days.

• If your issue is a duplicate, please close it and 👍 the existing issue instead
• To prevent auto-closure, add a comment or 👎 this comment

🤖 Generated with Claude Code

No this is a specific bug for macOS 26.2 update - that happened afterwards

https://developer.apple.com/documentation/Xcode-Release-Notes/xcode-26_2-release-notes

it started working - dupes are gone

i had it again today.

recorded a video about it
https://x.com/i/status/2019299768195850550

Update from today (2026-02-11): still happening on Claude Code v2.1.39, macOS 26.2 (25C56), zsh.

It's now showing both:

• [1] + suspended (signal)
• [1] + suspended (tty input)

Interactive TUI repeatedly redraws the splash + prompt block (same frame over and over), then suspends. --print mode is still fine.

I verified this is not MCP setup and reproduced after isolating startup variables:

1. Disabled all user plugins
2. Uninstalled failing plugins (rust-analyzer-lsp, graphyn-core)
3. Launched with --disable-slash-commands --no-chrome
4. Launched with strict empty MCP config (--strict-mcp-config --mcp-config /tmp/empty-mcp.json)

Even with all of that, the TUI still re-renders the same initial frame in a loop and suspends.

Also seeing recurring renderer diagnostics in debug logs consistent with prior reports, e.g. repeated high write ratios / write-only frames in earlier sessions (blit=0, write=...).

So at least in my environment, this reproduces even after removing plugin + MCP factors; looks like a core TUI/job-control interaction on macOS 26.2.

Additional datapoint from a fresh repro in v2.1.39 (debug enabled):

• During one redraw-loop window, debug log shows an auth event storm: 127 OAuth token check starting events in the same second (2026-02-11T10:47:44Z).
• This was observed after disabling/removing plugin/MCP variables as noted above.

I can share the full debug file if you want a direct attachment path/format.

<img width="774" height="867" alt="Image" src="https://github.com/user-attachments/assets/9dac0052-19e2-4758-bee1-44f04c464d9b" />

this is how it looks like

Fresh datapoint: full wipe + reinstall still reproduces on macOS 26.2 with v2.1.39.

What I did:

• Removed ~/.claude, ~/.claude.json, ~/.local/share/claude, ~/.local/bin/claude (clean reinstall state)
• Reinstalled via official installer (curl -fsSL https://claude.ai/install.sh | bash)
• No plugins installed (clean default)

Observed behavior in first launch:
1) Theme/onboarding screens repeatedly redraw/restart
2) Login flow completes successfully
3) Reaches normal prompt (Try "how do I log an error?")
4) Header/prompt redraw loop appears again
5) process stops with zsh: suspended (signal)

So this is reproducible after a true clean reinstall and before custom plugin/MCP restoration.

Possible repro clue from environment diff:

On affected machine session, shell env includes:

• TERM_PROGRAM=WarpTerminal
• LANG=C.UTF-8
• LC_ALL=C.UTF-8
• NO_COLOR=1

I can still repro the redraw/suspend path in this environment.

Suggest checking whether the TUI renderer behavior changes with locale/color env normalization (e.g. LANG/LC_ALL=en_US.UTF-8, NO_COLOR unset), since this machine differs from my other Mac where issue does not occur.

Latest update (workaround found on affected Mac):

A shell wrapper that launches interactive Claude inside a nested PTY (/usr/bin/script) appears to stop the recurring suspended (signal) / suspended (tty input) behavior in my Warp + macOS 26.2 environment.

Working wrapper pattern:

• normalize env (LANG/LC_ALL=en_US.UTF-8, unset NO_COLOR)
• for interactive runs, execute Claude via: script -q /dev/null ~/.local/bin/claude ...

Non-interactive modes (--print, --version, etc.) still run directly.

This looks like a TTY/job-control interaction specific to this environment, not install corruption (clean reinstall previously reproduced the issue).

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.