Codex app only displays (Something went wrong…)

Potential duplicates detected. Please review them and close your issue if it is a duplicate.

• #29304

Powered by Codex Action

I've run into the same problem. Has yours been resolved?

I ran into the same problem, and reinstalling it didn't work either

I've run into the same problem. Has yours been resolved?

No, the app is still broken for me.

> I've run into the same problem. Has yours been resolved? No, the app is still broken for me.

> I've run into the same problem. Has yours been resolved? No, the app is still broken for me.

I tried uninstalling and reinstalling, didn't work

the same to me, display "something wrong" after update. restart \ reinstall did't work.

我也是这个问题 重装没用

Installing the older version worked. It was likely a conflict caused by the new version.

Installing the older version worked. It was likely a conflict caused by the new version.

How did you install a previous version?

> Installing the older version worked. It was likely a conflict caused by the new version. How did you install a previous version?

wangnov/codex-app-mirror

I ran into the same issue, but I was able to use the Codex CLI; however, I tried getting it to fix the problem without success.

> > Installing the older version worked. It was likely a conflict caused by the new version. > > > How did you install a previous version? wangnov/codex-app-mirror

It didn't work; I installed the older version (OpenAI.Codex_26.616.4196.0_x64__2p2nqsd0c76g0.Msix), but I'm still getting the "something went wrong" error. Could there be an issue with my account?

> > > Installing the older version worked. It was likely a conflict caused by the new version. > > > > > > How did you install a previous version? > > > wangnov/codex-app-mirror It didn't work; I installed the older version (OpenAI.Codex_26.616.4196.0_x64__2p2nqsd0c76g0.Msix), but I'm still getting the "something went wrong" error. Could there be an issue with my account?

If you want to go back to a previous version, do not pick any version that starts with 26.616.

I went back to version 26.611 to get it working again.

Yeah, back to 26.611 it's ok

版本

yes, I install 26.611, it works.

> 版本 yes, I install 26.611, it works.是的，我安装了26.611，能用。

OpenAI.Codex_26.611.8604.0_x64__2p2nqsd0c76g0.Msix

> > > Installing the older version worked. It was likely a conflict caused by the new version. > > > > > > How did you install a previous version? > > > wangnov/codex-app-mirror It didn't work; I installed the older version (OpenAI.Codex_26.616.4196.0_x64__2p2nqsd0c76g0.Msix), but I'm still getting the "something went wrong" error. Could there be an issue with my account?

use this version OpenAI.Codex_26.611.8604.0_x64__2p2nqsd0c76g0.Msix

The issue still exists even after today's update.

reinstalling using Wangnov/Codex-App-Manager helped me.

I also ran into this issue. I had Codex extract the installation package for me, then loaded it locally instead of going through Microsoft’s program entry point.

I’m seeing the same issue on Windows.

After the latest update, Codex works normally at first, but after restarting the computer the app shows the same
“Something went wrong…” screen again. So, the problem comes back after a system restart.

I can reproduce a very similar failure on Windows/MSIX in the same 26.616 line, with one extra data point: launching the unpacked Codex.exe works, while Start menu / MSIX activation fails.

Environment

• Installed package: OpenAI.Codex_26.616.10790.0_x64__2p2nqsd0c76g0
• Previously working package: OpenAI.Codex_26.611.8604.0_x64__2p2nqsd0c76g0
• Windows: Windows 10 Pro, build 22621, HAL 10.0.22621.2506, x64
• Get-AppxPackage status for the installed package: Ok
• Direct unpacked launch works from: %USERPROFILE%\Downloads\OpenAI.Codex_26.616.10790.0_x64__2p2nqsd0c76g0\app\Codex.exe
• Start menu / MSIX activation fails and only shows Something went wrong

Important observation

The package contents do not appear corrupted. I compared the extracted app and the installed WindowsApps copy:

• app\resources\app.asar SHA256 is identical in both locations: B1445DE6388CAB35B3C1662646FC4231E63DBB1A9B08668260145B1BAA367C26
• app\Codex.exe SHA256 is identical in both locations: 9FA3C69CBA3DACC3BF70E5A824014F2899DFB6778DBE14724A1B815DBD94E32D
• app\resources\codex.exe SHA256 is identical in both locations: 7C8531062AA5B8252FD205852F7015412AAD646ABF4BE9C5CD5823EFE4836C71

So the difference seems to be MSIX/AppX activation context vs direct Win32 launch, not file corruption.

Relevant log excerpt

Log location:

%LOCALAPPDATA%\Packages\OpenAI.Codex_2p2nqsd0c76g0\LocalCache\Local\Codex\Logs\2026\06\25\codex-desktop-...-3472-t0-i1-051748-0.log

Key lines:

Launching app buildFlavor=prod packaged=true platform=win32

[AppServerConnection] Current reported app-server version: currentVersion=0.142.0 hostId=local
[AppServerConnection] initialize_handshake_result durationMs=233 initializeRequestId=__codex_initialize__ outcome=success transportKind=stdio

[window-manager] Failed to load bundle at errorCode=ERR_FAILED errorMessage="ERR_FAILED (-2) loading 'app://-/index.html'" errorName=Error
indexPath="C:\Program Files\WindowsApps\OpenAI.Codex_26.616.10790.0_x64__2p2nqsd0c76g0\app\resources\app.asar\webview\index.html"

[window-manager] window ready-to-show appearance=primary hostId=local startupElapsedMs=389 webContentsId=1 windowId=1

There is also a repeated warning before that:

bundled_executable_relocation_failed destinationPath=%LOCALAPPDATA%\OpenAI\Codex\bin\38dff8711e296435 errorCode=EPERM errorMessage="Bundled executable relocation failed during rename_staging" executableName=codex.exe operation=rename_staging platform=win32 sourceKind=windowsapps

However, I do not think that EPERM rename_staging is the direct trigger. The older working 26.611.8604.0 MSIX logs also contained similar bundled_executable_relocation_failed warnings, but the UI bundle loaded successfully. In 26.616, the consistent direct failure is ERR_FAILED (-2) loading 'app://-/index.html' from the WindowsApps/AppX launch path.

Current workaround

Launching the unpacked app\Codex.exe directly works normally and can open existing Codex sessions. Rolling back to 26.611.8604.0 also avoids the issue.

I have now updated to version 26.623.5175.0 and still have the same problem

<img width="2267" height="1460" alt="Image" src="https://github.com/user-attachments/assets/afbb6c9c-c145-4894-b57d-39968ee64ef4" />

I am seeing what looks like the same issue on Windows.

My environment:

• OS: Windows 11 x64
• Subscription: ChatGPT Plus
• Codex App package: OpenAI.Codex_26.623.8305.0_x64__2p2nqsd0c76g0
• Bundled app-server / Codex CLI version reported in logs: 0.142.3
• Standalone Codex CLI works normally on the same machine.

This still happens after uninstalling and reinstalling the Windows Codex App. It also happens on the first launch after reinstall, so it does not look like stale local user state.

The bundled app-server initializes successfully:

[StdioConnection] stdio_transport_spawned ... app\resources\codex.exe
[AppServerConnection] Transport start success
[AppServerConnection] Current reported app-server version: currentVersion=0.142.3
[AppServerConnection] initialize_handshake_result ... outcome=success
[AppServerConnection] Codex CLI initialized

But the renderer bundle fails to load:

[window-manager] Failed to load bundle
errorCode=ERR_FAILED
errorMessage="ERR_FAILED (-2) loading 'app://-/index.html'"
indexPath="C:\Program Files\WindowsApps\OpenAI.Codex_26.623.8305.0_x64__2p2nqsd0c76g0\app\resources\app.asar\webview\index.html"

I also verified that:

• app\resources\app.asar exists.
• webview\index.html exists inside app.asar.

So in my case, the file appears to exist, but Electron still fails to load it through the internal app://-/index.html route.

This looks like an Electron renderer / custom protocol / app:// route loading issue rather than a Codex CLI, network, or missing-file issue.

I'm using version 26.611.8604.0 and curious if this "Something went wrong..." bug has been fixed in the latest versions?

Did updating to a newer version resolve the problem for you? Should I upgrade from my current version to fix this?

I found a Windows workaround.

This does not appear to be caused by a missing webview/index.html. In my logs, the file exists, but Electron fails to load it through the custom protocol:

app://-/index.html

The failing path was:

...\resources\app.asar\webview\index.html

What fixed it for me:

1. Copy the installed Codex app folder to a normal writable location.
2. Rename the original resources\app.asar to resources\app.asar.original.
3. Extract app.asar into a real directory named resources\app.asar.
4. Launch app\Codex.exe from the copied folder.

The directory name matters. If I extract it to resources\app, the app starts as Codex (Dev) because the production metadata is not found. When the extracted directory is named app.asar, the app still reads resources\app.asar\package.json, keeps buildFlavor=prod, and webview/index.html is served from a normal NTFS directory instead of inside the asar archive.

After this workaround, my logs show:

buildFlavor=prod
Codex CLI initialized
window ready-to-show
[startup][renderer] app routes mounted

And the original error is gone:

ERR_FAILED (-2) loading 'app://-/index.html'

So this looks like an Electron/custom-protocol/asar loading issue in the Windows package, not an account, network, or CLI issue.

This is only a local workaround, not a proper fix, because the unpacked app runs outside the MSIX package and the updater is not available there.

I'm using version 26.611.8604.0 and curious if this "Something went wrong..." bug has been fixed in the latest versions? Did updating to a newer version resolve the problem for you? Should I upgrade from my current version to fix this?

After updating to version 26.623.81905 released on 7.1, this issue has been resolved for me.

After updating to the latest version, it works now.

Issue has been resolved in version 26.623.81905.