Skip to content

[BUG] Claude Code Desktop v1.1.8359 crashes within 90 seconds on first launch after auto-update, killing active MCP session #38393

@DC-AAIA

Description

@DC-AAIA

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?

When Claude Desktop auto-updates and restarts into a new version, the new version crashes within approximately 90 seconds of its first launch. The crash fires a beforeQuit handler with no user action and no preceding error in the main log. On second launch of the same version, the app runs stably. The first-boot crash interrupts any in-progress sessions, causing data loss and MCP server disconnects.

App crashes within 90 seconds of first post-update launch. main.log shows clean beforeQuit with no preceding error — the crash leaves no stack trace, suggesting a silent exception during first-boot initialization.

What Should Happen?

The new version should complete first-boot initialization and maintain session stability. If first-boot setup tasks require additional time, the app should not accept user sessions until initialization is complete.

Error Messages/Logs

2026-03-24 072305 info beforeQuitForUpdate handler fired, going down for update
2026-03-24 072337 info Starting app appVersion 1.1.8359 ...
[MCP servers connect successfully]
2026-03-24 072552 info beforeQuit handler fired, going down
[No preceding error — silent first-boot crash]

2026-03-24 075956 info Starting app appVersion 1.1.8359 ...
[Second launch — stable]

Steps to Reproduce

  1. Have Claude Desktop running an active session (e.g., a Claude Code session with GitHub MCP connected)

  2. A background auto-update completes and triggers a restart via beforeQuitForUpdate

  3. The new version launches and initializes all MCP servers successfully

  4. Within 60–90 seconds, beforeQuit fires with no user action — app exits

  5. Any in-progress GitHub MCP operations (e.g., create_or_update_file commits) are abandoned mid-operation

  6. Second launch of same version succeeds and runs stably

Claude Model

Opus

Is this a regression?

Yes, this worked in a previous version

Last Working Version

No response

Claude Code Version

2.1.78

Platform

Claude Code Desktop — Windows 10 x64

Operating System

Windows

Terminal/Shell

Windows Terminal

Additional Information

When Claude Desktop auto-updates and restarts into a new version, the new version crashes within approximately 90 seconds of its first launch. The crash fires a beforeQuit handler with no user action and no preceding error in the main log. On second launch of the same version, the app runs stably. The first-boot crash interrupts any in-progress sessions, causing data loss and MCP server disconnects.
To Reproduce

Have Claude Desktop running an active session (e.g., a Claude Code session with GitHub MCP connected)

A background auto-update completes and triggers a restart via beforeQuitForUpdate

The new version launches and initializes all MCP servers successfully

Within 60–90 seconds, beforeQuit fires with no user action — app exits

Any in-progress GitHub MCP operations (e.g., create_or_update_file commits) are abandoned mid-operation

Second launch of same version succeeds and runs stably

Expected behavior
The new version should complete first-boot initialization and maintain session stability. If first-boot setup tasks require additional time, the app should not accept user sessions until initialization is complete.
Actual behavior
App crashes within 90 seconds of first post-update launch. main.log shows clean beforeQuit with no preceding error — the crash leaves no stack trace, suggesting a silent exception during first-boot initialization.
Environment

OS: Windows 10/11 x64

Claude Desktop version at crash: 1.1.8359 (first launch post-update)

Previous stable version: 1.1.7714

Claude Code version: 2.1.78

Node version: 24.13.0

MCP servers active at time of crash: GitHub (v0.6.2), File System, n8n MCP Docs, n8n-mcp, railway-mcp-server

Log Evidence (main.log)
2026-03-24 072305 info beforeQuitForUpdate handler fired, going down for update
2026-03-24 072337 info Starting app appVersion 1.1.8359 ...
[MCP servers connect successfully]
2026-03-24 072552 info beforeQuit handler fired, going down
[No preceding error — silent first-boot crash]

2026-03-24 075956 info Starting app appVersion 1.1.8359 ...
[Second launch — stable]

Additional Context
The update to v1.1.8359 was downloaded on March 23, 2026 at approximately 4:37 PM CDT. The first-boot crash on March 24 at 7:25 AM CDT interrupted an active GitHub MCP session that was performing multi-file commits, resulting in a partial push and orphaned session state. The second boot at 7:59 AM ran stably with no changes made by the user.
This appears to be a race condition between first-boot initialization tasks (session migration, plugin synthesis, cache rebuild) and session acceptance. The fix should gate session availability until first-boot initialization is confirmed complete.

Metadata

Metadata

Assignees

No one assigned

    Labels

    invalidIssue doesn't seem to be related to Claude Code

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions