[Bug] Android Remote became unusable: Windows host appears disconnected and long tasks do not open
What version of the Codex App are you using (From “About Codex” dialog)?
ChatGPT/Codex Desktop for Windows 26.814.5167.0; bundled codex-cli 0.148.0-alpha.15.
Android controller: latest ChatGPT build offered by Google Play on 2026-08-21; Google Play reports no update available. The exact Android build, OS version, and device model were not captured yet.
What subscription do you have?
Paid ChatGPT account (exact tier intentionally omitted).
What platform is your computer?
Host: Microsoft Windows NT 10.0.26200.0 x64
Controller: ChatGPT for Android.
What issue are you seeing?
Codex Remote from ChatGPT Android became effectively unusable over approximately the last 24 hours after working reliably about 48 hours ago.
There are two reproducible failure modes:
- In most attempts, Remote does not connect to the Windows host at all. The host is shown with a red/offline indicator even though ChatGPT/Codex Desktop is open, awake, online, and signed in. Android displays:
Failed to connect to ChatGPT desktop
Make sure the ChatGPT desktop app is open, then try again.
- When Remote does connect, it lists the tasks currently running on the Windows host. Short/small tasks open normally, but selecting a long task does not open the transcript and eventually shows the host/task as disconnected. This makes the tasks that most need remote supervision inaccessible.
The regression began around 2026-08-20. Other users known to the reporter independently describe the same recent Android Remote failure. The OpenAI status page showed no Remote Connections incident when checked on August 20 and 21.
What steps can reproduce the bug?
- Keep ChatGPT/Codex Desktop
26.814.5167.0open on an online Windows host with Remote enabled. - Open ChatGPT for Android, using the latest build available from Google Play.
- Enter Remote and select the paired Windows host.
- Frequently, observe the red/offline host indicator and the
Failed to connect to ChatGPT desktoperror despite the Desktop app being open. - Retry until the connection succeeds and the host task list appears.
- Open a small/recent task; it normally loads.
- Open a long-running task with substantial history; it remains unavailable or transitions to disconnected instead of rendering the transcript.
What is the expected behavior?
Android Remote should reliably connect to an awake, online, paired Windows host. After connecting, every visible task should open at least a bounded recent transcript tail without requiring the complete history to hydrate first. Long tasks must not block or tear down the host connection.
Additional information
A screenshot captured around 2026-08-21T14:28Z shows the paired Windows host offline and the exact connection error above. The host name, account information, task names, IDs, and contents are omitted from this public report. A sanitized screenshot and diagnostics can be provided privately.
Related issue #38023 documents Android Remote failing to hydrate an idle large task and a turn/start timeout. The long-task symptom overlaps with that report. This issue is distinct because the primary regression frequently occurs before any task is selected: Android marks the Windows host disconnected and cannot enter Remote at all.
Related issue #37126 documented old Android chats not opening, but the reporter closed it after adopting shorter chats as a workaround rather than after a confirmed product fix.
11 Comments
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
there appear to be lots of duplicates because this issue hasn't even been looked at by the dev team because I'm hitting it too daily until last night when it's stopped working completely and now even the new chat takes 30 seconds to connect
Devs please! How can you ignore a remote connection bug. Please fix this.
Same here. Duplicate or no, this needs attention
Yes, started happening in the last few days, incredibly annoying.
The app clearly gets SOME form of connection sometimes, because I can see the updated thread list and start new threads, but since yesterday or the day before, thread loading from the app is like 95% failure rate.
My theory is that it's doing something with the full chat history because it blew out hundreds of MBs of mobile data as it just sat spinning...
I was able to reproduce my issue and it matches https://github.com/openai/codex/issues/39916 from my desktop logs have codex look into it for you for each individual issue. There are a ton to go around though 8)
Same..
I'm having the same issue and can reproduce it with the same steps on the latest builds.
same
Same here for 2 days
same for 2 days