yeachan-heo/oh-my-codex

OMX, short for oh-my-codex, is a workflow layer that sits on top of the OpenAI Codex CLI, a command-line coding assistant from OpenAI. The README is clear that OMX does not replace Codex: Codex still does the actual agent work, while OMX wraps it with stronger defaults, reusable workflows, and persistent state so longer tasks are easier to manage. In practice, after installing both the Codex CLI and OMX through npm, you launch a session with a single command and drive your work using keyword shortcuts inside Codex. The README highlights a handful: $deep-interview for clarifying a request, $ralplan for proposing and approving a plan, $ralph for carrying an approved plan through to completion, $team for splitting work across parallel executors, and $ultragoal for stitching multiple Codex goals into a durable launch. Plans, logs, memory, and runtime state are stored in a project-local .omx/ directory so progress survives across sessions. A doctor command checks the install is healthy, and an exec command proves Codex can actually authenticate and complete a model call. You would reach for OMX if you already use Codex regularly and want more structure around it, repeatable roles, clearer hand-offs between planning and execution, and a way to coordinate parallel work or persistent loops. If you prefer plain Codex with no extra layer, the README says you probably do not need OMX. The project is written in TypeScript, requires Node.js 20 or newer, and the recommended environment is macOS or Linux with tmux available; native Windows is supported but explicitly less polished. The full README is longer than what was provided.