AI CODING TOOLS / LESSON 13
What is Codex?
Meet OpenAI's coding agent and the places you can use it
Loading the lesson…
Loading activities…
Browse lessons
AI CODING TOOLS / LESSON 13
Meet OpenAI's coding agent and the places you can use it
Loading the lesson…
Loading activities…
You just spent a whole stack of lessons getting cozy with Claude Code. Now let's meet its cousin. Codex is an AI coding agent made by OpenAI, the company behind ChatGPT. Same as Claude Code, it can read your project, change files, run commands and explain what it did.
By the end of this lesson you'll know what Codex is, where you can use it, what it's good at and how it stacks up next to Claude Code.
Codex is a coding agent, and that word "agent" is doing real work. A chatbot answers questions. An agent actually DOES stuff.
When you hand Codex a task, here's what it can do.
You describe what you want in plain English, same as you've been doing with Claude Code, and Codex figures out the steps.
Codex shows up in a few different places, and they all share the same brain. Pick whichever one feels most comfy.
The CLI is a program you run in your terminal by typing codex. It works right on the files sitting on your computer. This is the closest match to Claude Code, and it's the one we'll set up next lesson.
Fun fact for later... the Codex CLI is open source. Its code is public on GitHub at github.com/openai/codex.
If you like living inside a code editor, Codex has an extension for VS Code and the editors built on it, like Cursor and Windsurf. JetBrains IDEs and Xcode have their own ways to pick Codex as the assistant. The extension sits right beside your code, so you can point at the file you've got open and ask about it.
The ChatGPT desktop app (for Mac and Windows, with a Linux version in preview) has Codex built right in. You open a folder, pick Codex instead of regular ChatGPT and give it a task. Great for longer jobs, and great for folks who'd rather not stare at a terminal.
Codex Cloud runs tasks on OpenAI's computers instead of yours. You connect a GitHub repository, set up an environment once, and then hand it tasks from the web, your phone or the desktop app. It keeps chugging along even while your computer is asleep. When it's done, you review the changes and open a pull request if you like what you see.
Codex handles the same kinds of everyday coding work you've watched Claude Code do.
/review command that looks over your changes and reports problems it finds, without editing anything.It's got some safety habits baked in, too. On your own computer, Codex runs inside a sandbox, which is a fenced-off area that limits what it can touch. By default it can only change files in your project folder, and it has no internet access unless you say so. We'll dig into that more in the setup lesson.
Here's the simple version. Claude Code and Codex are the same kind of tool, made by two different companies.
Both are AI coding agents. Both read your project, make changes, run commands and ask your permission (each in its own way). Both can live in your terminal, your code editor, a desktop app or the web.
They also both read a special instructions file you keep in your project. Claude Code uses CLAUDE.md, which you met earlier in this module. Codex uses AGENTS.md. Same idea, different name. And here's the good news... Claude Code can read an AGENTS.md file too, which makes running both on one project a whole lot easier. There's an entire lesson on that coming up.
Say you open a project folder and start Codex. You type this.
Codex reads through the folder and answers with a quick tour. Then maybe you ask this.
Codex finds the form, makes the change, runs your tests and tells you what it did. You check the change before you keep it. If that feels familiar, it should, because it's the exact same loop you've been running with Claude Code.
CLAUDE.md and AGENTS.md)Next lesson you'll install the Codex CLI, sign in and give it its very first safe task in a real project folder...