Skip to content

missing context % usage in new version #18101

Closed

Description

@MauScheff

What version of Codex CLI is running?

codex-cli 0.120.0

What subscription do you have?

API Key:

Which model were you using?

gpt-codex-5.4.

What platform is your computer?

Darwin 25.3.0 arm64 arm

What terminal emulator and version are you using (if applicable)?

ghostty

What issue are you seeing?

Since a recent upgrade, I no longer see the % left of the context window. I noticed that this may be intentional as we now have decent auto-compaction, but in the cli I like to have more control myself rather than this being abstracted away. Sometimes the agent does start acting "drunk" and I rather have control over this as to continue like that, or start a new session. I sometimes rather do either my own /compact when I want, or even instructing the agent to write a Handoff.md for the next agent, but I prefer to do this myself and not rely on auto-compaction.
Having more control, is the reason I prefer the cli over codex desktop too. Thank you for your consideration.

What steps can reproduce the bug?

% usage / tokens left in context window is not longer shown.

What is the expected behavior?

I can see the % usage as before

Additional information

As a cli user I like to have more control over the "raw machine", not too much, like we had to decide before the temperature, but with the token window I rather not auto handle that.

Activity

  1. added
    CLIIssues related to the Codex CLI
    contextIssues related to context management (including compaction)
    on Apr 16, 2026
  2. etraut-openai commented on Apr 16, 2026

    @etraut-openai
    Contributor

    This will be addressed in the next release.

  3. MauScheff commented on Apr 16, 2026

    @MauScheff
    Author

    thanks!

  4. MauScheff commented on May 13, 2026

    @MauScheff
    Author

    I would partly attribute this to my latest unexpected $6,000 bill. I had no idea how much context I was using and growing / shrinking and just kept going because gpt.5.5 is really good at this (auto-compaction) so I kept reusing the same thread with old and unnecessary context over and over. I think having this indicator we had would have nudged me to my older approach where I would manually start new threads as the context was getting too big, or even just seeing how fast it grew. I see that in the new version it's disabled by default and you have to look for it and enable it in the status line settings (I now did), I think this is a dark UX pattern and you should always provide some kind of in-app gas meter by default, but maybe it's not deliberate, that's why I'm saying this. I'm a solopreneur and I'm having a significant struggle with this unexpected bill.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    CLIIssues related to the Codex CLIbugSomething isn't workingcontextIssues related to context management (including compaction)

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions