Skip to main content
Back

oh-my-codex Hits 32K Stars Wrapping OpenAI Codex in Multi-Agent Power

Yeachan-Heo released OMX, a workflow layer that wraps OpenAI's Codex CLI with agent teams, hooks, git-worktree isolation, and persistent state.

Repo, Full Access
  • Yeachan-Heo released oh-my-codex (OMX), an MIT-licensed workflow layer wrapping OpenAI's Codex CLI.
  • Ships 33 role prompts, 37 skills, and 5 MCP servers for durable state, memory, and code intelligence.
  • Team mode gives every worker its own git worktree by default, with automatic leader merge integration.
  • Supports mixed-provider teams running Codex, Claude, and Gemini workers side by side via env config.
  • Canonical flow is $deep-interview then $ralplan then $ultragoal, keeping planning and execution phases separate.
  • Install via npm; macOS and Linux with tmux are the tuned path, Windows via WSL2.

A new open source project called oh-my-codex, shortened to OMX, has crossed 32,000 stars on GitHub as a runtime layer that sits on top of OpenAI's Codex CLI. Rather than replacing Codex, it adds the plumbing that most people end up rebuilding themselves once they start running Codex against real repositories: coordinated multi-agent teams, git worktree isolation, persistent plans and logs, native hooks, and a HUD for watching what the agent is doing.

The pitch from creator Yeachan Heo is deliberately narrow. OMX does not replace Codex; it adds a better working layer around it, where Codex does the actual agent work and OMX provides role keywords, reusable skills, and a durable `.omx/` folder for plans, logs, memory, and runtime state. If you already use Codex CLI happily and just want a nicer harness around it, this is aimed squarely at you.

The layer around the model

OMX is a TypeScript project distributed through npm. Install is a one-liner alongside Codex itself, and the recommended launch flags turn on stronger reasoning and bypass the usual approval prompts for trusted repos.

npm install -g @openai/codex
npm install -g oh-my-codex
omx setup && omx doctor
omx --worktree=feat/task --madmax --xhigh

The `--madmax` flag is OMX shorthand for Codex's `--dangerously-bypass-approvals-and-sandbox`, which removes the normal approval and sandbox guardrails, so it should only be used in trusted repositories. `--high` and `--xhigh` are shorthand for setting model reasoning effort, and when using `--madmax` from a git repository the recommended pattern is a worktree launch rather than running directly in the current checkout.