
What is Codex?
Codex is currently OpenAI's coding agent: you describe what you want done in words, Codex reads the source code in your project directory, edits files, runs commands, and presents the changes for you to review. According to the README of the github.com/openai/codex repository (read on 03/11/2026), Codex CLI runs on your computer, is open source under the Apache-2.0 license, and comes with IDE extensions, desktop applications, and cloud-based Codex. This post does not promise that Codex always writes correctly, does not compare benchmarks, and does not specify which ChatGPT plan OPCAIOS uses. The post covers: how Codex differs from the old 2021 Codex model, four places to run it, login, pricing read from the official page today, risks, a brief comparison with Claude Code, when to use it, and OPCAIOS's real experience.

One agent, four places to run, same loop: give a task, Codex works in a sandbox, you review the diff. Diagram: OPCAIOS.
How is today's Codex different from the 2021 Codex model?
Today's Codex is a complete coding agent, while 2021's Codex was just a code generation model. The paper "Evaluating Large Language Models Trained on Code" by Mark Chen and colleagues (arXiv 2107.03374, submitted 07/07/2021, read on 03/11/2026) describes Codex at that time as a GPT model fine-tuned on public source code from GitHub, and a production version of this model powered GitHub Copilot. The old model received a description and returned a code snippet; it did not open files on its own, did not run commands, did not work with git. Current Codex is software that runs on your machine or on OpenAI's cloud, using newer models under the hood, reads your project on its own, edits multiple files, runs tests, and reports results. When reading guides online, check the publication date to avoid mixing up the two. If you're not sure what an agent is, see What is an AI agent.
Where does Codex run?
Codex runs in four places, using the same account. The table below is compiled from the README and Codex documentation at learn.chatgpt.com/docs (developers.openai.com/codex redirects here, read on 03/11/2026):
| Place to run | How to open | Good for |
|---|---|---|
| Codex CLI | Type codex in terminal in your project directory |
People comfortable with terminal, needing automation with codex exec |
| IDE extension | Extensions for VS Code, Cursor, Windsurf; Xcode and JetBrains have their own integrations | People wanting to see diff beside code |
| Desktop application | ChatGPT app for macOS, Windows, select Codex; from CLI type codex app |
People who prefer graphical interface |
| Cloud-based Codex | ChatGPT on web, mobile or desktop, select cloud environment and GitHub repository | Long tasks, running even when your machine is off |
Current documentation calls the desktop application "ChatGPT desktop app", with Codex being one mode within it. Cloud-based Codex gives each task its own workspace; you view changes, ask for further edits, then commit or open a pull request (see Codex on the cloud). With CLI, version 0.162.1 on our machine lists subcommands like exec, review, resume, mcp, cloud, app, and doctor; installation instructions are in Installing Codex CLI.
Log in to Codex with ChatGPT plan or API key?
There are two ways to log in to Codex: a ChatGPT account to use your subscription plan's quota, or an API key to pay per usage. According to the Authentication page in Codex documentation (read on 03/11/2026), desktop application, Codex CLI, and IDE extensions accept both methods when working on your machine, but cloud-based Codex only accepts ChatGPT login. In CLI, codex login opens your browser to log into ChatGPT; an API key is passed through standard input with codex login --with-api-key; codex login status tells you which method you're using. With an API key, costs are calculated by standard API pricing on your OpenAI Platform account and some cloud features are not available. The documentation recommends using an API key for automated processes like CI. Login information is saved on your machine, by default in a plain text file at ~/.codex/auth.json or in your operating system's credential store, so do not share or push this file to git. An account logging in with email and password must enable multi-factor authentication to use cloud-based Codex.
How much does Codex cost today?
Codex is not sold separately: Codex comes in ChatGPT plans, or billed via API key. Codex's pricing page (developers.openai.com/codex/pricing redirects to learn.chatgpt.com/docs/pricing, read 03/11/2026), USD per month, before tax, subject to change:
| Plan | Price | What the pricing page says about Codex |
|---|---|---|
| Free | 0 | One standard-speed model in the desktop app, varies by rollout |
| Go | 8 | Like Free |
| Plus | 20 | Web, CLI, IDE plugins, iOS; automatic code review, Slack; buy extra credits |
| Pro | 100, 200 or 500 | Everything in Plus; currently no 5-hour window limit |
| Business | 20/person per year or 25 per month | ChatGPT and Codex, larger instances for cloud work |
| API Key | Per API pricing | CLI, SDK, IDE plugins; no cloud features |
The Plus limit is only stated as an estimate of messages per 5-hour window on device, varies widely by model, may have weekly limits too; cloud work consumes the limit faster. The feature table by plan does not list CLI for Free and Go, so if you want to be sure CLI works, count from Plus or an API key. This is not financial advice; the pricing table sits next to Claude Code and Cursor in the post Claude Code, Cursor or Codex CLI.
What are Codex's limits and risks?
Codex's risk comes from Codex editing files and running real commands on your machine. Per the "Agent approvals & security" page (read 03/11/2026), Codex on device runs in a sandbox enforced by the OS, usually can only write to the working directory, network is off by default, plus an approval policy that decides when to ask you. A git directory gets Auto mode suggested; a non-git directory gets read-only suggested; in the default write mode, .git and .codex are still protected read-only. Five things to remember:
- Sandbox bypass flag:
--dangerously-bypass-approvals-and-sandbox(alias--yolo) turns off both sandbox and approvals; only use in an isolated VM. - Prompt injection: content from the web or unfamiliar files may contain instructions that make the agent do wrong things.
- Wrong code that looks right: always read the diff and test-run it.
- Rate limit: long tasks hit the limit fast; check with
/status. - Secrets: do not paste keys, passwords into prompts; see security when using AI to write code.
How to pick a mode for each task is in the post Codex approval and sandbox modes.
How is Codex different from Claude Code?
Codex and Claude Code are alike in the way they work: assign jobs by talking, run in terminal, IDE, desktop and cloud. The main differences are the model provider (OpenAI vs Anthropic), how you pay (ChatGPT plans vs Claude plans), guide file (Codex reads AGENTS.md, Claude Code reads CLAUDE.md) and source code (Codex CLI is open under Apache-2.0). Full table in the post Claude Code, Cursor or Codex CLI; how to have both agents work on one project is in the post use Codex and Claude Code at the same time.
When should I use Codex?
Use Codex when Codex fits the plan you are paying for and the way you work. Sensible cases:
- Already paying ChatGPT Plus or higher: Codex comes with the plan, try it before buying more.
- Want to read the tool's source code: Codex CLI is open source.
- Need long-running work and opening pull requests: use Codex on cloud after linking GitHub.
- Need to automate with scripts: see codex exec run automation.
- Want a second reviewer:
codex reviewand/reviewread changes without editing the work tree.
Not yet if you have no git or backup. The process for newcomers is in the post programming agent for people who don't code.
OPCAIOS Experience
The main machine of OPCAIOS runs Codex CLI version 0.162.x and the Codex desktop application, used in parallel with Claude Code and Google Antigravity; we don't use Cursor daily. What is real:
- AGENTS.md set at user level for Codex, content equivalent to Claude Code's file: reply in Vietnamese, read documentation in order at session start, write logs at session end.
- config.toml sets the default model, thinking level "medium", adds one model provider, subagent uses a smaller model, and marks project directories as trusted.
- Each task assigned from the desktop application has its own working directory by date.
- Multiple agents coordinate through a shared state file and daily log, each session commits only its own files. Once encountered over 1,500 "other" files just because of line break style when one agent deployed from a Windows machine.
We haven't used Codex in the cloud to open pull requests, haven't run codex exec in CI, don't have command-blocking hooks (hooks currently only report session status), and haven't measured Codex quality or cost against other agents.
Frequently Asked Questions
Is Codex free?
Partly. According to the pricing page read on 03/11/2026, the Free 0 USD plan includes one standard speed model in the desktop application, subject to rollout, and the feature table does not list CLI for this plan. To use CLI, IDE utilities, and Codex in the cloud, the pricing page starts at Plus 20 USD/month; or use an API key and pay per usage. Prices do not include tax and may change.
Is Codex the same as GitHub Copilot?
No. In 2021, one production version of the old Codex model ran GitHub Copilot, according to arXiv paper 2107.03374. Today's Codex is OpenAI's separate agent product, logged in with ChatGPT or API key, runs in terminal, IDE, desktop application, and cloud. Copilot is GitHub's product with its own plans and usage.
Can I use Codex if I don't know how to code?
You can for small tasks if you keep safety discipline. Start on a trial project with git, assign only read tasks first (for example "explain this project"), then move to fixing one small change and read the diff before accepting. When you don't understand the diff, ask Codex to explain each change in plain language.
Does Codex read CLAUDE.md files?
No, by default. According to the AGENTS.md page in the Codex documentation (read on 03/11/2026), Codex looks for AGENTS.override.md, then AGENTS.md, then alternative names you declare in project_doc_fallback_filenames. The documentation also has an /import command to import settings from Claude Code, but we haven't tried it. The simple way is to put common rules in AGENTS.md; see What is AGENTS.md.
Where should I learn Codex?
Follow this order: read the Codex CLI page and Authentication page in the official documentation, install CLI per the README, run a read-only task on a trial project, practice reading /diff, then learn approval mode, AGENTS.md, and automation. If you want someone to guide you through your actual work, see OPCAIOS's AI application training course.
Source: github.com/openai/codex (README), learn.chatgpt.com/docs (Codex CLI, Codex IDE extension, ChatGPT desktop app, Codex Cloud, Authentication, Agent approvals & security, AGENTS.md, Pricing; developers.openai.com/codex redirects here), arXiv 2107.03374 (Chen et al., 2021), codex --help version 0.162.1, read on 03/11/2026; prices do not include tax and may change; limits are only pricing page estimates; CLI usage with Free and Go plans not verified; /import command not tried. Experience is OPCAIOS's.
Translated from Vietnamese with AI. Read the original.
Want to apply this in your business?
Book a free 30-minute call (Vietnam time, in English or Vietnamese).
