claude-code - 💡(How to fix) Fix [BUG] [Critical] Claude Desktop v1.8555.2 crashes on launch with V8 OOM on macOS Tahoe 26.4.1

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?

Hi Anthropic Support,

Claude Desktop is completely unusable on my Mac due to a reproducible V8 out-of-memory crash on launch. After ~6 seconds the app crashes with SIGTRAP every time, before I can interact with the UI in any meaningful way. I've performed a full clean reinstall — the issue persists.

— Environment —————————————————————————————————————————— App version: Claude.app 1.8555.2 (downloaded from claude.com/download) macOS: 26.4.1 (build 25E253) — Tahoe Hardware: MacBook Air, Apple M3, 24 GB RAM Account: Claude Max Reproducibility: 100% (4 consecutive crashes, fresh reinstall included)

— Root cause (confirmed via --enable-logging=stderr) ———————— V8 old-space hits its default heap ceiling at ~2807–3166 MB during startup. Last-resort Mark-Compact GC frees zero bytes, then OOM aborts the renderer.

Excerpt from the log (one representative crash):

<--- Last few GCs ---> Mark-Compact (reduce) 3166.7 (3322.6) -> 3166.7 (3251.1) MB ... last resort; GC in old space requested Mark-Compact (reduce) 3166.7 (3251.1) -> 3166.7 (3249.8) MB ... last resort; GC in old space requested

ERROR:electron/shell/common/node_bindings.cc:185 OOM error in V8: CALL_AND_RETRY_LAST Allocation failed - JavaScript heap out of memory

zsh: trace trap

The startup work appears to load far too many heavy resources concurrently (Cowork rootfs.img.zst ~10GB download, Claude Code binary, 16 skills, 30 chat conversations, MCP server lists, GitHub repos, multiple telemetry SDKs — Datadog RUM, Sentry, GTM, FB Pixel), all kept live in the same V8 heap.

— Workarounds attempted, all FAILED ————————————————————

1. NODE_OPTIONS="--max-old-space-size=12288" → Ignored. Log shows: "WARNING:electron/shell/common/node_bindings.cc:502 NODE_OPTIONS ignored due to disabled nodeOptions fuse."
2. Tried 8192 MB and 12288 MB via env var and launchctl setenv — both ignored.
3. --disable-gpu flag → rejected: "GPU access not allowed."
4. Full state wipe: deleted ~/Library/Application Support/Claude, all Caches, Saved Application State, HTTPStorages, WebKit, Preferences plist, Cookies binarycookies. Reset TCC for Documents/Downloads/Desktop.
5. Complete uninstall + fresh download from claude.com/download + reinstall — identical crash.

— Request ——————————————————————————————————————————— Please either: (a) Raise V8 max-old-space-size in the packaged Electron config (e.g. to 8 GB). (b) Re-enable the nodeOptions Electron fuse so users can configure it via environment variable. (c) Reduce concurrent startup work — defer Cowork VM download, skill loading, and telemetry SDK initialization until after the main UI has stabilized.

This appears related to GitHub issue #48276 (reported crashes on macOS 26.3.1+) and may affect all Tahoe users with sufficient account state (skills, projects, chat history) to push V8 over the default heap ceiling on startup.

Happy to provide the full --enable-logging=stderr output (~50 MB) on request.

In the meantime I'm using claude.ai in the browser, which works fine.

Thanks, Jorvik

What Should Happen?

fix

Error Messages/Logs

Steps to Reproduce

Hi Anthropic Support,

Claude Desktop is completely unusable on my Mac due to a reproducible V8 out-of-memory crash on launch. After ~6 seconds the app crashes with SIGTRAP every time, before I can interact with the UI in any meaningful way. I've performed a full clean reinstall — the issue persists.

— Environment —————————————————————————————————————————— App version: Claude.app 1.8555.2 (downloaded from claude.com/download) macOS: 26.4.1 (build 25E253) — Tahoe Hardware: MacBook Air, Apple M3, 24 GB RAM Account: Claude Max Reproducibility: 100% (4 consecutive crashes, fresh reinstall included)

— Root cause (confirmed via --enable-logging=stderr) ———————— V8 old-space hits its default heap ceiling at ~2807–3166 MB during startup. Last-resort Mark-Compact GC frees zero bytes, then OOM aborts the renderer.

Excerpt from the log (one representative crash):

<--- Last few GCs ---> Mark-Compact (reduce) 3166.7 (3322.6) -> 3166.7 (3251.1) MB ... last resort; GC in old space requested Mark-Compact (reduce) 3166.7 (3251.1) -> 3166.7 (3249.8) MB ... last resort; GC in old space requested

ERROR:electron/shell/common/node_bindings.cc:185 OOM error in V8: CALL_AND_RETRY_LAST Allocation failed - JavaScript heap out of memory

zsh: trace trap

The startup work appears to load far too many heavy resources concurrently (Cowork rootfs.img.zst ~10GB download, Claude Code binary, 16 skills, 30 chat conversations, MCP server lists, GitHub repos, multiple telemetry SDKs — Datadog RUM, Sentry, GTM, FB Pixel), all kept live in the same V8 heap.

— Workarounds attempted, all FAILED ————————————————————

1. NODE_OPTIONS="--max-old-space-size=12288" → Ignored. Log shows: "WARNING:electron/shell/common/node_bindings.cc:502 NODE_OPTIONS ignored due to disabled nodeOptions fuse."
2. Tried 8192 MB and 12288 MB via env var and launchctl setenv — both ignored.
3. --disable-gpu flag → rejected: "GPU access not allowed."
4. Full state wipe: deleted ~/Library/Application Support/Claude, all Caches, Saved Application State, HTTPStorages, WebKit, Preferences plist, Cookies binarycookies. Reset TCC for Documents/Downloads/Desktop.
5. Complete uninstall + fresh download from claude.com/download + reinstall — identical crash.

— Request ——————————————————————————————————————————— Please either: (a) Raise V8 max-old-space-size in the packaged Electron config (e.g. to 8 GB). (b) Re-enable the nodeOptions Electron fuse so users can configure it via environment variable. (c) Reduce concurrent startup work — defer Cowork VM download, skill loading, and telemetry SDK initialization until after the main UI has stabilized.

This appears related to GitHub issue #48276 (reported crashes on macOS 26.3.1+) and may affect all Tahoe users with sufficient account state (skills, projects, chat history) to push V8 over the default heap ceiling on startup.

Happy to provide the full --enable-logging=stderr output (~50 MB) on request.

In the meantime I'm using claude.ai in the browser, which works fine.

Thanks, Jorvik

Claude Model

None

Is this a regression?

Yes, this worked in a previous version

Last Working Version

No response

Claude Code Version

1.81

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

No response