You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Cloud environment publish fails, then setup chat cannot resume #51434
I tried to publish a Codex cloud environment from the Codex desktop app on Windows. Publishing failed, and every subsequent message in that environment’s setup chat now fails to resume.
Previous messages remain visible. Other cloud chats still work.
Errors
The Environment panel shows:
“Couldn’t publish environment. Your settings are saved. Try publishing again.”
Sending a message in the affected chat shows:
“Failed to resume chat. Something went wrong. Request ID: a2d8e97c-a1ae-4f1d-97ac-3dc91b0b66bd”
The setup assistant also reported that the cloud workspace could not start:
“failed to query executor configuration capabilities”
Steps leading to the issue
Worked in a cloud environment setup chat.
Attempted to publish the environment.
Publishing failed.
Tried sending additional messages, including “try again” and “hey.”
Each message failed to resume the chat.
Expected behavior
The environment should publish successfully, or a publishing failure should leave the setup chat usable so I can diagnose and retry it.
Additional context
The assistant said the environment was already published, but the Environment panel showed publication failed. The assistant also reported that the branch push was blocked because the workspace could not start. I cannot confirm whether the environment published or whether the latest code reached the remote repository.
Please investigate and help restore access while preserving saved environment settings and any unpushed work.
I encountered an environment publishing failure twice on 2026-10-10, with the same message reported above.
Both attempts used the Codex web UI in a signed-in remote cloud Chromium browser. I created two separate setup chats for the same existing saved environment and installed small Markdown instruction files plus a Python installation/validation script.
In each setup instance, installation completed, all eight tests passed, SHA-256 checks matched, and reinstalling was idempotent. The setup draft saved successfully.
Publish started at approximately 05:55 UTC and 07:15 UTC. Both attempts remained in Publishing for several minutes, then displayed:
“需要排查发布问题”
“无法确认你的设置是否已保存。请先报告此问题,再重新开始。”
Approximate translation: “Publishing needs investigation” and “We couldn’t confirm whether your settings were saved. Please report this issue first, then start again.”
Retry was disabled, and the saved environment’s published version ID did not change. After the first failure, reloading/checking status briefly showed Publishing again, then returned to the same error. Starting a fresh setup chat did not resolve it.
I don’t have an HTTP 500 trace or app-server logs from these attempts, so I can’t establish the cause.
What is the safe recovery path while preserving the existing environment and setup, and which diagnostic information would help investigate this?
Additional reproduction from a personal ChatGPT Plus account (October 12, 2026).
Clients: ChatGPT iPhone app and mobile Safari; personal workspace.
Observed:
Creating a Codex Cloud environment works on desktop.
Cloud environments and sessions fail to load on mobile, both in the iPhone app and mobile Safari (error/loading spinner).
GitHub repository selection works on mobile.
Attempting environment setup on mobile shows:
Error starting chat
Could not start cloud environment setup (unavailable)
Opening an existing cloud task created on desktop from iPhone displays "This task couldn't connect".
Network comparison: Same failure on Wi-Fi and cellular on the iPhone.
Other-service comparison: Claude cloud sessions work on this iPhone. This is not a controlled comparison of backends, but it indicates general cloud-session connectivity is available.
Expected: Cloud environments and existing tasks created on desktop should load and reconnect on mobile under the same personal Plus account.
Desktop creation succeeding, together with the same mobile failure across Wi-Fi/cellular and app/Safari, points toward a Codex-specific mobile environment-discovery/task-reconnection issue rather than a simple local network problem. Root cause is not confirmed.
Activity
github-actions commentedon Oct 6, 2026
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
philippdubach commentedon Oct 9, 2026
I'm running into this on macOS too, in the Codex desktop app. I can't publish the cloud environment or resume its setup chat.
This happened during setup after I changed Internet access to "All (unrestricted)". The assistant reported these errors:
The Environment panel says:
Retrying didn't help. I also got:
I submitted feedback through the app.
Could you check this macOS occurrence too and advise how to recover the saved setup?
ChrisTinaAmadeus commentedon Oct 9, 2026
我在 Windows 桌面端也遇到过:界面提示设置已保存,但 Publish 失败,对话一直加载。
10 月 1 日北京时间 22:30 左右,日志里发布完成接口两次返回 HTTP 500,同时出现
Codex app-server is not available和No active workspace。22:40 左右,保存草稿的请求也返回了 500。希望能明确哪些设置已保存,并提供保留环境和已有工作的恢复办法。
kjin2020 commentedon Oct 10, 2026
mine shows "Codex app-server is not available" and "publishing needs investigation. we couldn't confirm your setup was saved ...."
rhuang19 commentedon Oct 10, 2026
I encountered an environment publishing failure twice on 2026-10-10, with the same message reported above.
Both attempts used the Codex web UI in a signed-in remote cloud Chromium browser. I created two separate setup chats for the same existing saved environment and installed small Markdown instruction files plus a Python installation/validation script.
In each setup instance, installation completed, all eight tests passed, SHA-256 checks matched, and reinstalling was idempotent. The setup draft saved successfully.
Publish started at approximately 05:55 UTC and 07:15 UTC. Both attempts remained in Publishing for several minutes, then displayed:
“需要排查发布问题”
“无法确认你的设置是否已保存。请先报告此问题,再重新开始。”
Approximate translation: “Publishing needs investigation” and “We couldn’t confirm whether your settings were saved. Please report this issue first, then start again.”
Retry was disabled, and the saved environment’s published version ID did not change. After the first failure, reloading/checking status briefly showed Publishing again, then returned to the same error. Starting a fresh setup chat did not resolve it.
I don’t have an HTTP 500 trace or app-server logs from these attempts, so I can’t establish the cause.
What is the safe recovery path while preserving the existing environment and setup, and which diagnostic information would help investigate this?
JeffSteinbok commentedon Oct 13, 2026
Additional reproduction from a personal ChatGPT Plus account (October 12, 2026).
Clients: ChatGPT iPhone app and mobile Safari; personal workspace.
Observed:
Expected: Cloud environments and existing tasks created on desktop should load and reconnect on mobile under the same personal Plus account.
Desktop creation succeeding, together with the same mobile failure across Wi-Fi/cellular and app/Safari, points toward a Codex-specific mobile environment-discovery/task-reconnection issue rather than a simple local network problem. Root cause is not confirmed.