Please add a user-facing “Plan Mode” toggle similar to Claude’s. In Plan Mode, Codex should inspect the repo, ask clarifying questions if needed, and produce a proposed implementation/validation plan without editing files, running destructive actions, committing, or pushing. The user should then be able to approve the plan, request changes, or switch back to normal execution mode. This would be especially useful for coordinating multi-agent work, reviewing risky migrations, and separating design discussion from implementation.
Thanks for sharing this idea @jon.power, and welcome to the community.
Good news: a Plan Mode workflow is already available in Codex. You can instruct Codex to inspect the repository, analyze the task, ask clarifying questions, and produce a proposed implementation and validation plan before making any changes. This is especially useful for larger refactors, migrations, and multi-agent coordination.
The current approach is prompt-driven rather than a dedicated UI toggle, but it supports the same separation between planning and execution that you’re describing.
You can find more details and examples here:
https://developers.openai.com/codex/learn/best-practices
That said, a dedicated user-facing toggle could still make this workflow more discoverable and easier to use, so it’s a reasonable feature request. Curious to hear how others are handling planning vs. execution in their Codex workflows.
-Mark G.