VS Code extension regression: restore collapsible “Working” (hide step-by-step logs, show final answer)

Open 💬 0 comments Opened Jan 22, 2026 by MaxMiksa

What version of the VS Code extension are you using?

VS Code plugin “Codex - OpenAI's coding agent” 0.4.62

What subscription do you have?

Via API.

Which IDE are you using?

Visual Studio Code

What platform is your computer?

Microsoft Windows NT 10.0.26200.0 x64

What issue are you seeing?

In a previous stable version of the extension, each assistant turn had a “Working” header that could be clicked to collapse all intermediate records (thinking/actions/tool logs), leaving only the clean final response visible. This made long tasks much easier to read.

In recent versions (0.4.62), the process/log output appears to be always expanded (or no longer collapsible). For multi-step tasks this creates very long “Working” sections and makes it hard to quickly find the final answer; it requires a lot of scrolling.

Related discussion: https://github.com/openai/codex/discussions/9603

What steps can reproduce the bug?

  1. Install/update the VS Code extension to 0.4.62.
  2. Run a prompt that triggers multiple steps/tool calls so the “Working” section becomes long.
  3. Try to collapse the per-turn “Working” section.

Observed: the “Working” section cannot be collapsed (or is always expanded).

What is the expected behavior?

  • Restore click-to-collapse/expand for “Working” per turn.
  • (Optional) Provide a setting to default-collapse “Working” (show final output by default, expand on demand).

Additional information

  • Codex: OpenAI Codex v0.87.0
  • Model: gpt-5.2 (medium)

View original on GitHub ↗