@matifali PR ready for review

coder/coder#19118

@hugodutka @DevelopmentCats, please review and let me know if this needs any changes :)

@hugodutka thanks for the review, committed the suggested changes.

@hugodutka @matifali made the changes.

@35C4n0r when I ran the module in my workspace I got the following error:

time=2025-08-08T16:37:35.443Z level=INFO msg="Starting server on port" port=3284
agent exited with error:
    github.com/coder/agentapi/cmd/server.runServer
        /home/runner/work/agentapi/agentapi/cmd/server/server.go:129
  - ========
    error: unexpected argument '--skip-git-repo-check' found       
                                                                   
      tip: to pass '--skip-git-repo-check' as a value, use '-- --sk
    ip-git-repo-check'                                             
                                                                   
    Usage: codex-aarch64-unknown-linux-musl [OPTIONS] [PROMPT]     
           codex-aarch64-unknown-linux-musl [OPTIONS] [PROMPT] <COM
    MAND>                                                          
                                                                   
    For more information, try '--help'.
    ========:
    github.com/coder/agentapi/cmd/server.runServer.func1
        /home/runner/work/agentapi/agentapi/cmd/server/server.go:115
  - non-zero exit code:
    github.com/coder/agentapi/lib/termexec.init
        <autogenerated>:1

Codex has probably been updated since you last tested if fully. Could we add a test to check that codex works end to end, meaning that it starts successfully after installing the latest version?

@hugodutka will do so, looks like there have been 9 minor releases since I raised this PR, I developed in 0.10.0 now it 0.19.0

@hugodutka how should we handle this, should we give a base minimum to maximum version ? They haven't released a stable/major version yet, and will continue to push breaking changes regularly.

I'd just add a test that installs the latest version and checks if it starts. The test will notify us if they push a breaking change later.

The test will notify us if they push a breaking change later.

makes sense, the tests fail now

@hugodutka added the test.

0_19_0.patch

These are the minimal changes that would be needed to support 0.19.0 (maybe with tuning the prompt for task_reporting a bit), but the TUI have been restructured https://github.com/openai/codex/releases/tag/rust-v0.11.0 since 0.11.0.
We can take this up as a separate issue/PR