Codex suits teams already paying for ChatGPT that want a coding agent in the terminal, IDE, ChatGPT desktop app and cloud, with many agents running in parallel and automatic pull request reviews. Cursor suits developers who want an AI-first code editor where the agent, chat and completions live inside the place they write code, with a choice of models. Codex is included in every ChatGPT plan, including Free; Cursor has its own Free plan with limited requests.
Both tools now do far more than autocomplete: they take a task, change files and check the result. I read OpenAI's Codex page and its help article on using Codex with a ChatGPT plan, plus Cursor's home, plans and docs pages.
Codex and Cursor head to head
| Codex | Cursor | |
|---|---|---|
| What you get | OpenAI's coding agent | AI coding agent and editor |
| Where you use it | ChatGPT desktop app, CLI, IDE extension, web | Cursor app, CLI, cloud, mobile |
| Free access | Included in ChatGPT Free and Go | Free plan with limited requests |
| Cloud agents | Codex Cloud on Plus, Pro and business plans | Cloud agents listed |
| Code review | Reviews pull requests automatically | AI code review listed |
| Models | OpenAI models | Several, including Grok models |
| Extending it | Plugins | Rules, MCP and Skills |
The plan detail that trips people up
OpenAI's help centre says Codex is included across ChatGPT plans, including Free and Go, but Codex Cloud is not. Cloud tasks need an eligible Plus, Pro, Business, Enterprise, Healthcare or Education account, and usage limits vary by plan. So a free user can run Codex locally in the CLI or IDE, but can't hand off a task to run in the cloud while the laptop is shut.
Another useful line: when you're signed in with ChatGPT, Codex's auto-review safety checks don't count toward your plan's usage limits.
Where Codex fits best
Codex is built for multi-agent work. OpenAI says it runs many agents in parallel, uses computer and browser tools to verify its work, keeps going in reusable cloud environments after you close your laptop, and reviews pull requests on its own. Its refreshed CLI adds an agent view, voice and worktrees, which helps if your developers live in the terminal.
Where Cursor fits best
Cursor is an editor first. The agent, chat and context-aware completions sit inside the app where you write and read code, and you can pick models from several providers, with third-party models charged separately by usage. Its docs cover Rules, MCP servers and Skills, so teams can teach it house conventions once and reuse them.
Another contender is explained in my guide to what Google Antigravity's free plan covers.
How to choose between them
- Check whether your team already has ChatGPT seats that include Codex.
- Decide if developers want a new editor or prefer their current one plus a CLI.
- List the tasks you'd hand to a cloud agent overnight.
- Check which models your company allows.
- Run the same real ticket through both and compare the pull requests.
For a step-by-step start, read my guide on how to use OpenAI Codex, and see my wider list of AI coding assistants for other options.
How each one fits a team's workflow
Codex's CLI is simple to start: you open your project folder, run codex and sign in with your ChatGPT account. OpenAI says plugins connect Codex to your other tools so it can act across your apps. Cursor's site lists agents, cloud, a CLI, mobile access, automations and code review, so it's built to follow a task from the editor to the pull request.
Not a developer? My Replit and Lovable comparison covers builders that need less code.
My advice on Codex or Cursor
If your company already pays for ChatGPT, start with Codex because it's included, and test a Codex Cloud task on a paid seat. If your developers want AI woven through the editor and a choice of models, give them Cursor's Free plan for a week. Plenty of teams end up with both: Cursor for writing code, Codex for parallel tasks and reviews.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
Do you make AI coding tools? Tool makers can ask to be considered for a featured place in comparisons like this one.
Frequently asked questions
Is Codex free?
Codex is included across ChatGPT plans, including Free and Go, with usage limits that vary by plan. Codex Cloud isn't included with Free or Go.
Is Cursor free?
Cursor lists a Free plan with limited agent and chat requests, alongside paid individual and Teams plans.
Can Codex run in VS Code?
OpenAI lists a Codex IDE extension, along with the Codex CLI, Codex web and Codex mode in the ChatGPT desktop app.
Which models does Cursor use?
Cursor offers models from several providers, including Grok models, with third-party models charged separately by usage.
Can Codex review pull requests?
Yes. OpenAI says Codex automatically reviews pull requests, surfaces findings and answers questions.
Do Codex safety checks use my limits?
No. OpenAI says auto-review safety checks are free and don't count toward your plan's usage limits when you're signed in with ChatGPT.
How do I start Codex in the terminal?
OpenAI says to open your project directory, run codex and sign in with your ChatGPT account when prompted.
