Yeachan-Heo/oh-my-codex

oh-my-codex (OmX) hit #1 on GitHub Trending with 2,852 stars in a single day. Total: 11,600+ stars. The reason is simple. It turns OpenAI's Codex CLI from a solo tool into a multi-agent orchestration system with teams, hooks, HUDs, and persistent memory. Think oh-my-zsh for AI coding agents. Codex CLI is the execution engine. OmX wraps it with structured workflows: $deep-interview for requirements clarification, $ralplan for architectural planning, $team for coordinated parallel execution, and $ralph for autonomous research. Each command triggers a specialized agent pipeline. The multi-agent execution model is the standout feature. Team workers get isolated git worktrees by default. That means multiple agents can modify the same codebase simultaneously without merge conflicts. Each worker operates in its own branch, in its own working directory, with its own context. A durable task graph manages dependencies, and a review queue lets you approve changes before they merge. Installation takes two commands: npm install -g @openai/codex oh-my-codex followed by omx setup. The tmux-first runtime means each agent gets its own terminal pane. You watch them work in real-time through the HUD. v0.11.12 (April 2, 2026) is the latest release. Persistent state lives in the .omx/ directory: plans, execution logs, memory across sessions. The intent-first interview system ($deep-interview) captures requirements before any code gets written, reducing wasted iterations. Built primarily in TypeScript (91.5%) with Rust (4.7%) for performance-critical paths. MIT licensed. Actively maintained with rapid release cadence. If you use Codex CLI and want to scale beyond single-agent workflows, OmX is the tool. For a comparison of AI coding IDEs, read Cursor vs Windsurf 2026. Explore more AI coding tools on Skila.