Windows Remote Control fails with false “only one instance” error

What version of the Codex App are you using (From “About Codex” dialog)?

Windows app/package version: 26.903.8094.0 Log-reported release: 26.901.51231

What subscription do you have?

Pro

What platform is your computer?

OS: Microsoft Windows NT 10.0.26200.0 Architecture: x64 Codex/ChatGPT Windows package: 26.903.8094.0 Log-reported release: 26.901.51231

What issue are you seeing?

I’m having a persistent issue with ChatGPT/Codex Remote Control on Windows 11 Pro.

The app intermittently shows:

• “Couldn’t update remote control availability”
• “Could not enable remote control. Please ensure only one instance of ChatGPT is running”

I verified that only one actual ChatGPT application instance is running.

Diagnostics performed:

1. Fully exiting ChatGPT from the system tray leaves zero ChatGPT/Codex processes.

2. Launching ChatGPT once creates one root ChatGPT process tree plus normal child processes and one codex.exe backend.

3. Process inspection confirms there is not a second independent ChatGPT instance.

4. Remote Control still immediately reports the false “only one instance” error after a clean launch.

Relevant local state:

• .codex-global-state.json contains valid-looking Remote Control installation/environment IDs
• Remote migration/backfill flags are complete
• no obvious duplicate local app instance was found

Relevant log lines during failure include:

[remote-connections/window-context] refresh_remote_control_started previousConnectionCount=0

followed by:

refresh_remote_control_completed autoConnectConnectionCount=0 nextConnectionCount=0 previousConnectionCount=0

I also repeatedly see:

[devicecheck-attestation] DeviceCheck attestation unavailable deviceCheckLatencyMs=null durationMs=0 errorCode=1

The issue persists across full app exits/relaunches.

Environment:

• Windows 11 Pro
• ChatGPT/Codex Windows app
• Package path observed under:
  OpenAI.Codex_26.903.8094.0_x64__2p2nqsd0c76g0
• Remote access from mobile is affected
• The Windows PC itself remains reachable normally over Tailscale/RDP
• This is not a general network outage on the machine

This appears to be a Remote Control registration/enrollment/attestation issue rather than an actual duplicate-process issue.

Please let me know if there is a supported way to reset only the Remote Control enrollment/registration state without wiping local Work/Codex project state.

What steps can reproduce the bug?

.

What is the expected behavior?

.

Additional information

.

Start with the clean-launch process tree, .codex-global-state.json, and the remote-connections/window-context and devicecheck-attestation log lines. Compare the registration, enrollment, and attestation state with the false single-instance error; done means one Windows instance can enable Remote Control and allow mobile access without wiping Work/Codex state.