[BUG] Claude Desktop v1.3109.0 crashes on launch (V8 OOM) — macOS 26.5 beta
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?
<h2>Environment</h2>
Detail | Value
-- | --
App version | 1.3109.0
macOS | 26.5 beta (Build 25F5042g)
Hardware | Mac Mini, Apple Silicon M2, 24 GB RAM
<h2>Description</h2>
<p>Claude Desktop crashes with a V8 out-of-memory error every time the app launches Code mode. Regular chat mode works fine. The crash occurs within ~8-12 seconds of the VM initializing, consistently hitting ~3.2-3.3 GB of V8 heap before dying.</p>
<p>This is a <strong>fully clean install</strong> — all app data, caches, MCP servers, IndexedDB, and vm_bundles were deleted before reproducing. The crash occurs with zero configuration.</p>
<h2>Additional Observations</h2>
<ul>
<li><strong>Regular chat mode works</strong> — the app is stable as long as Code/Cowork is not activated</li>
<li><strong><code>NODE_OPTIONS="--max-old-space-size=6144"</code> has no effect</strong> — Electron appears to ignore the flag; heap still caps at ~3.2 GB</li>
<li><strong><code>BuddyBleTransport</code> errors fire on every launch</strong> regardless of crash, possibly a Bluetooth companion feature incompatible with macOS 26.5 beta APIs</li>
<li><strong><code>MaxListenersExceededWarning</code></strong> appears on every launch — potential EventEmitter memory leak contributing to the OOM</li>
<li><strong>IndexedDB LevelDB lock errors</strong> also appear on launch (<code>LOCK: File currently in use</code>) but are non-fatal</li>
<li><strong>vm_bundles had ballooned to 7.9 GB</strong> on the previous install before cleanup; the fresh install begins rebuilding the VM bundle (<code>sessiondata.img created</code>) immediately before crashing</li>
</ul>
<h2>Troubleshooting Already Attempted</h2>
<p>All of the following were tried with no improvement to Code/Cowork stability:</p>
<ul>
<li>Full app reinstall (twice — including deleting /Applications/Claude.app and redownloading)</li>
<li>Deleted all caches: Cache/, Code Cache/, GPUCache/, DawnGraphiteCache/, DawnWebGPUCache/, Crashpad/</li>
<li>Deleted vm_bundles/ (was 7.9 GB)</li>
<li>Deleted IndexedDB and removed stale LOCK file</li>
<li>Removed all MCP server configuration (claude_desktop_config.json set to empty mcpServers)</li>
<li>Deleted Preferences, Saved Application State, WebKit, and Caches plist files</li>
<li>Set <code>NODE_OPTIONS="--max-old-space-size=6144"</code> before launch</li>
<li>80%+ system memory free at time of crash</li>
</ul>
<h2>Related Issues</h2>
<ul>
<li>#48261 — V8 OOM / JavaScript heap out of memory (macOS 26.3.1, M5)</li>
<li>#32170 — Claude Desktop crashing on opening (macOS 26.3, M4)</li>
<li>#29625 — Crashes on launch (macOS 26.4 beta, M2)</li>
<li>#26646 — Freezing, API 500 errors, 12GB VM bundle (macOS 26.3)</li>
</ul>
What Should Happen?
<p>Code and Cowork modes should launch without exhausting the V8 heap.</p></body></html>
Error Messages/Logs
(node:23687) [DEP0040] DeprecationWarning: The `punycode` module is deprecated.
(node:23687) MaxListenersExceededWarning: Possible EventEmitter memory leak detected.
11 change listeners added to [RAr]. MaxListeners is 10.
Error occurred in handler for '$eipc_message$_..._claude.buddy_$_BuddyBleTransport_$_reportState':
Error: No handler registered for '..._BuddyBleTransport_$_reportState'
<--- Last few GCs --->
[23687:0x10400500000] 5940 ms: Mark-Compact (reduce) 3180.0 (3299.0) -> 3180.0 (3217.3) MB,
pooled: 0.0 MB, 13.28 / 0.00 ms (average mu = 0.562, current mu = 0.322) last resort;
GC in old space requested
[23687:0x10400500000] 5954 ms: Mark-Compact (reduce) 3180.0 (3217.3) -> 3180.0 (3207.8) MB,
pooled: 0.0 MB, 13.66 / 0.00 ms (average mu = 0.381, current mu = 0.001) last resort;
GC in old space requested
[23687:0416/174420.707885:ERROR:electron/shell/common/node_bindings.cc:185]
OOM error in V8: CALL_AND_RETRY_LAST Allocation failed - JavaScript heap out of memory
Steps to Reproduce
<ol>
<li>Fresh install of Claude Desktop 1.3109.0 from claude.ai/download</li>
<li>Delete all app data:
<pre><code>rm -rf ~/Library/Application\ Support/Clauderm -rf ~/Library/Caches/com.anthropic.claudefordesktoprm -rf ~/Library/Preferences/com.anthropic.claudefordesktop
</code></pre></li>
<li>Launch from Terminal: <code>/Applications/Claude.app/Contents/MacOS/Claude</code></li>
<li>Regular chat loads successfully</li>
<li>Navigate to Code or Cowork → crash within seconds</li>
</ol>
Claude Model
None
Is this a regression?
Yes, this worked in a previous version
Last Working Version
_No response_
Claude Code Version
v1.3109.0
Platform
Anthropic API
Operating System
macOS
Terminal/Shell
Terminal.app (macOS)
Additional Information
_No response_
3 Comments
Found 3 possible duplicate issues:
This issue will be automatically closed as a duplicate in 3 days.
🤖 Generated with Claude Code
Resolved — root cause was transcript bloat in
~/.claude/Posting the fix in case anyone else hits the same silent Code tab crash.
Environment
~/.claude/had grown to 13 GB from ~1 week of continuous Claude Code sessions on a single projectDiagnosis
All 10 oversized files belonged to one project. Top offenders:
Notably, the subagent
agent-acompact-*.jsonlfiles under a single session folder added up to several GB on their own. The bloat is not just in the top-level transcript.Fix
Archived the entire project folder (rather than picking individual
.jsonlfiles) to keep parent sessions and their subagent folders together:Result:
13 GB → 74 MB. Code tab loaded instantly on next launch. No reinstall, no cache clear, no other changes needed.
Tip for anyone hitting this
If you've been running Claude Code on the same project continuously for days or weeks, you're a strong candidate for this crash. Start fresh sessions periodically rather than resuming the same one indefinitely — the on-disk codebase is the real context, not the transcript history.
Product suggestion
A pre-load size check with a clear "session too large — archive older transcripts" message would save a lot of head-scratching. Silent crashes with no diagnostic output are brutal for non-technical users who won't think to launch from Terminal to capture stderr.
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.