Feature request: add setting to control command/tool output folding

Potential duplicates detected. Please review them and close your issue if it is a duplicate.

• #18508

Powered by Codex Action

Potential duplicates detected. Please review them and close your issue if it is a duplicate. * Codex desktop app shows verbose tool-trace rows in chat and hurts readability #18508 _Powered by Codex Action_

Thanks for the duplicate check. These issues are related, but I think this request is slightly different.

#18508 / #18396 focus more on hiding or collapsing verbose tool-trace rows, while this issue is specifically asking for a Codex Desktop preference to control command/tool output folding, including an "always expand command/tool outputs" option and/or remembering the user's expand/collapse choice.

The motivation is debugging/test workflows where command output is the main evidence, and repeatedly expanding every command block adds friction. Related CLI-side request: #4550, but this one is Desktop-specific.

yes please like verbose mode in claude code

+1 from a macOS Codex Desktop power-user workflow.

This is the main thing that keeps the app from feeling like a practical replacement for the CLI for me. In the CLI, I can supervise the work live because shell commands, tool calls, and their output are visible as Codex runs. In the Desktop app, those details are hidden behind small expand controls, and I often have to expand repeatedly within a single session, sometimes even for one tool/action flow.

The important part is not just final-summary detail. I want a durable inspection preference such as:

• always expand command/tool calls and stdout/stderr by default
• remember my expand/collapse choice per thread/session
• keep the current clean collapsed UI for users who prefer it

This matters for active supervision: tests, builds, linters, debugging output, browser/tool captures, and command failures are the evidence I use to decide whether Codex is headed in the right direction. Waiting until the final answer makes it much harder to steer long-running coding work.

Related desktop visibility issues show the same underlying need from slightly different angles: #16415 and #19891.

I found myself getting the same headaches from this. In my opinion claude code handles this correctly and its annoying that OpenAI still havent added such a simple feature.

I made an unofficial open-source workaround that adds two Appearance preferences:

• Expand tool output
• Show full shell commands

https://github.com/BJDubb/codex-full-output

The project installs a separate patched copy and leaves the Microsoft Store installation untouched. It is currently Windows-only and tested against the versions documented in the repository.
I’d still prefer these controls to be supported natively but I’m sharing the workaround here for anyone affected in the meantime.

<img width="1624" height="1264" alt="Image" src="https://github.com/user-attachments/assets/98132ffa-eab0-4bb2-a210-29b3fcb59c10" />
<img width="1806" height="884" alt="Image" src="https://github.com/user-attachments/assets/b7f8e329-7a5d-4014-8bf6-20b5df953ad7" />

This is still a major problem in Codex Desktop for macOS 26.803.41515 (6321), and the current UI now has two levels of disclosure.

Tool activity is first collapsed into a one-line aggregate summary such as “loaded tools / ran commands / searched the web”. Expanding that summary only reveals truncated one-line command rows. Each row must then be opened again to inspect stdout/stderr in a smaller secondary panel.

This makes long debugging and audit-heavy sessions extremely difficult to review while scrolling through the thread. The problem is therefore not only that output is collapsed by default: it is nested collapse, forced one-line truncation, and a reduced output surface.

Please add a persistent Classic / Inspection mode that:

• shows command invocations directly in the chronological transcript;
• shows stdout/stderr inline in a normal full-width code block;
• keeps tool and command output expanded by default;
• remembers the preference globally or per thread;
• may cap very large outputs with an inline scroll area, but does not reduce every command to one line.

The clean compact mode can remain available, but power users need an opt-out from the nested, TLDR-style presentation. The existing “Steps with code output” level does not restore the previous inspectable command display.

Related: #16415 and #19891.