Switching one thread to Full Access appears to change all active threads across projects

Resolved 💬 2 comments Opened Mar 29, 2026 by lastobelus Closed Jul 2, 2026

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

26.325.31654

What subscription do you have?

ChatGPT Pro

What platform is your computer?

Darwin 25.4.0 arm64 arm

What issue are you seeing?

Switching one thread to Full Access appears to change all active threads across projects to Full Access.

What steps can reproduce the bug?

  1. Open multiple active threads in the Codex desktop app.
  2. Make sure at least one other thread is open in a different project.
  3. Leave the threads on Custom (config.toml).
  4. In one thread, change the permissions setting to Full Access.
  5. Return to the other active thread(s), including threads in other projects, and run a command that would previously have been governed by Custom (config.toml).

What is the expected behavior?

I expect the change to apply only to the thread where I switched from Custom (config.toml) to Full Access, unless the UI clearly says (and warns before changing) that the change is global.

What I see instead:

Changing one thread from Custom (config.toml) to Full Access appears to change every active thread to full access, including threads in other projects.

Additional information

I ran into this while trying to work around #10390, which also affects me. I normally use Custom (config.toml). After changing a single thread to Full Access, all other threads also reported being "Full Access". Not just threads in the current project, but all threads across allprojects, even currently active threads doing unrelated work.

I may be misunderstanding the intended model, but if this is expected behavior, I think the UI should state clearly that the setting is global rather than thread-local or project-local, and provide a clear warning or confirmation dialog before making such a global change.

View original on GitHub ↗

This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗