OPENAI'S CODING AGENT
Codex is OpenAI's coding agent. Like Claude Code, it runs in your terminal, sees the project folder you start it in, and can read files, write code, and run commands — with your approval.
If you've done the Claude Code course, you already understand Codex. The shape is the same: an AI that acts on your project instead of just talking about it. What changes is the company behind it, the command you type, and some of the vocabulary.
That's genuinely worth noticing. These tools are converging, so the habits you build transfer. Learn to give a clear instruction and to review a change carefully, and you can pick up any of them in an afternoon.
Codex was originally the name of an AI model, not a tool. Around 2021, OpenAI trained a version of GPT specifically on source code, and called it Codex. It could turn a comment like “sort this list by date” into working code — which, at the time, felt like witchcraft.
That model became the engine inside GitHub Copilot, the autocomplete that finishes your line of code as you type. For years, that was the shape of AI coding: a very good autocomplete sitting inside your editor.
Then the ambition grew. Autocomplete helps you type; it doesn't do the job. So OpenAI brought the name back for something bigger: an agent — a Codex that doesn't just suggest the next line, but takes a whole task, works through the files, runs the tests, and comes back with a finished change for you to review.
So when you hear "Codex," ask which one is meant. The old model is history. The Codex you install today is the agent.
The name “Codex” has meant two different things. What does it mean today?
Codex and Claude Code are made by the same company.