Segfault crash during MCP reconnect via /mcp dialog (Bun 1.3.11)

Status Open
Reported on v2.1.87
Maintainer reply None cached
Activity 0 comments · opened Jul 31, 2026

Bug Report

Claude Code version: 2.1.87
Bun version (embedded): 1.3.11
Platform: macOS Silicon (Darwin 25.5.0, macOS 26.5.2)

What happened

Claude Code crashed with a segmentation fault while the user was reconnecting an MCP server (xray-bridge, SSE transport on localhost:8443) via the /mcp dialog. The user had just clicked "Enable" on the xray-bridge entry when the crash occurred within approximately one second.

Context leading up to crash

  1. Mid-conversation, the xray-bridge MCP server had disconnected (the host machine had been rebooted)
  2. Claude suggested reconnecting via /mcp
  3. User opened /mcp, disabled xray-bridge, then clicked Enable
  4. Crash occurred during or immediately after the enable action

Crash details from terminal output

Bun v1.3.11 (759ce802) macOS Silicon
macOS v26.5.2
CPU: fp aes crc32 atomics
Args: "claude"

Elapsed: 151855187ms | User: 632794ms | Sys: 64298ms
RSS: 0.45GB | Peak: 0.54GB | Commit: 0.02ZB | Faults: 833

panic: Segmentation fault at address 0x0
oh no: Bun has crashed. This indicates a bug in Bun, not your code.

Crash report link: https://bun.report/1.3.11/M_1759ce80mgkgEuhogC_uxkvQuxkvQmmp2lB215qgB__A2AA

Notes

  • Session had been running for ~1.75 days (151,855,187ms elapsed)
  • The Commit: 0.02ZB value appears to be a display bug
  • The null pointer dereference (address 0x0) suggests a use-after-free or uninitialised pointer during the SSE reconnection flow
  • All conversation state was lost — no graceful recovery
  • Two installs present: standalone binary at ~/.local/bin/claude (v2.1.87) and npm/Homebrew at v2.1.89 — neither auto-updated

Expected behaviour

MCP reconnection via /mcp should not crash the process. If the SSE connection fails, it should surface an error message, not segfault.

View original on GitHub ↗