What carries straight over
Here's the good news, itemized. Almost everything you learned in the Claude CLI chapter is true of the Codex CLI too. You are not starting from zero. You're starting from "most of the way there."
- It's a terminal tool. You run it from the same terminal you set
up two chapters ago. Same window, same prompt, same
cdto get into the right folder first. - It works in the folder you start it in. Start Codex inside a folder and that folder is its world for the session. It isn't roaming your whole machine.
- You instruct it in plain English. No special code, no flag soup. You describe the outcome you want, the way you'd brief a capable coworker. The briefing skill from the Claude chapter is the same skill here.
- It runs the agent loop. Look, plan, act, check, repeat. Codex reads, plans the next step, does it, checks the result, and goes again until the job is done or it needs you. Same engine.
- It asks before it acts. When Codex wants to change a file or run a command, it stops and shows you the action first. Reading that line before you approve is the same habit, and it's still the most important one.
- You sign in with an account, in a browser. No API key, no credit card for normal interactive use. Same OAuth shape as the Claude sign-in.
If you understood the Claude CLI chapter, you already understand how the Codex CLI behaves. That's not a simplification. The two tools genuinely converged on the same design because it's the design that works.
What's left is the differences, and they're smaller than this list. The next step covers them.
What carries straight over
Here's the good news, itemized. Almost everything you learned in the Claude CLI chapter is true of the Codex CLI too. You are not starting from zero. You're starting from "most of the way there."
- It's a terminal tool. You run it from the same terminal you set
up two chapters ago. Same window, same prompt, same
cdto get into the right folder first. - It works in the folder you start it in. Start Codex inside a folder and that folder is its world for the session. It isn't roaming your whole machine.
- You instruct it in plain English. No special code, no flag soup. You describe the outcome you want, the way you'd brief a capable coworker. The briefing skill from the Claude chapter is the same skill here.
- It runs the agent loop. Look, plan, act, check, repeat. Codex reads, plans the next step, does it, checks the result, and goes again until the job is done or it needs you. Same engine.
- It asks before it acts. When Codex wants to change a file or run a command, it stops and shows you the action first. Reading that line before you approve is the same habit, and it's still the most important one.
- You sign in with an account, in a browser. No API key, no credit card for normal interactive use. Same OAuth shape as the Claude sign-in.
If you understood the Claude CLI chapter, you already understand how the Codex CLI behaves. That's not a simplification. The two tools genuinely converged on the same design because it's the design that works.
What's left is the differences, and they're smaller than this list. The next step covers them.