[BUG] Desktop app crashes (GPU process gone) when the browser pane loads login/auth pages
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?
The Claude Desktop app on Windows crashes completely (the whole window closes,
no error dialog) when the built-in browser pane loads authentication/login
pages — e.g. Hostinger's hPanel login, or a Keycloak OIDC login page.
Reproduced consistently on two different Windows machines.
In %APPDATA%\Claude\logs\main.log every crash ends with:
GPU process gone: { type: 'GPU', reason: 'crashed', exitCode: 101457950, serviceName: 'GPU' }
and the log cuts off immediately — no graceful shutdown lines, no Crashpad
report is written.
Notable pattern: in all 8 occurrences across two days, the crash is preceded
(~1 second earlier) by this warning:
[PreviewContext] Blocked subresource to private-resolving host { resourceType: 'xhr' }
8 out of 8 crashes show that exact warning right before dying, so it looks
correlated with the private-resolving-host subresource blocking path in the
browser pane, triggered by auth pages (redirect chains, beacons, probes).
What Should Happen?
The login page should load in the browser pane (or fail gracefully with an
error inside the pane). The desktop app should never close entirely, even if
the page's GPU/rendering work or a blocked subresource misbehaves.
Error Messages/Logs
2026-08-17 04:17:30 [warn] [PreviewContext] Blocked subresource to private-resolving host { resourceType: 'xhr' }
2026-08-17 04:17:31 [info] GPU process gone: {
type: 'GPU',
reason: 'crashed',
exitCode: 101457950,
serviceName: 'GPU'
}
(log cuts off here — next line is the app starting again 15 minutes later)
Same pattern on 2026-08-16: seven "GPU process gone" crashes between 01:12
and 02:12, every one preceded ~1s by the same [PreviewContext] warning.
Steps to Reproduce
- Start a Claude Code session in the desktop app.
- Open the built-in browser pane and navigate to a login page
(e.g. https://hpanel.hostinger.com or any Keycloak login screen).
- Within a few seconds the entire app closes.
Claude Model
None
Is this a regression?
Yes, this worked in a previous version
Last Working Version
Claude Desktop 1.30096.5 (MSIX / Microsoft Store install — note the log says "CCD-autoupdate Disabled: MSIX install"), Windows 11 Home build 26200, x64. Reproduced on a second Windows 11 machine as well. One of the two machines also logs kernel video-engine timeouts (WER Kernel_141), but the other reproduces the crash without any known GPU/driver issues.
Claude Code Version
Fable 5 (claude-fable-5)
Platform
Anthropic API
Operating System
macOS
Terminal/Shell
Terminal.app (macOS)
Additional Information
MSIX / Microsoft Store install — startup log says "CCD-autoupdate Disabled:
MSIX install", so the app cannot self-update. Windows 11 Home build 26200, x64.
Reproduced on a second Windows 11 machine (crashes on Hostinger hPanel
authentication). One of the two machines also logged kernel video-engine
timeouts (WER Kernel_141) and WHEA events (Kernel_124) around the crashes;
the other machine reproduces without known GPU issues.
This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗