Codex by OpenAI: parallel coding agents bundled with every ChatGPT plan
Updated on September 6, 2026
Codex is OpenAI's software engineering agent, built to write code, fix bugs and review pull requests on its own. There is no separate Codex subscription, your ChatGPT plan is your Codex plan, free tier included. The same agent runs in the terminal, in VS Code, on the web and on iOS, with an open source CLI for local work. Paid tiers simply raise the ceiling.
- Free trial access with no card required
- Several background tasks running in parallel
- Automatic pull request review on GitHub
- Open source CLI for local sessions
- Credit consumption is hard to predict on big jobs
- Five-hour rolling quota drains fast in intense sessions
- Cloud features reserved for paid tiers
Hand Codex a ticket, get back a pull request
Codex runs complete software tasks in isolated cloud containers with your GitHub repository preloaded. You assign a bug, a feature or a test pass; it reads the codebase, runs linters and tests, then returns a documented pull request with logs attached. Three jobs running side by side on a Monday morning reshapes the whole week.
Every session lives in a sandbox) enforced at the kernel level, and you can redirect the agent mid-task without losing context. Under the hood, the GPT-5.6 family does the heavy lifting, from the high-end Sol to the thrifty Luna, heirs to the February 2026 GPT-5.3-Codex milestone that OpenAI describes as the first model instrumental in building itself (a line that made the rounds among engineers).
From the terminal to the subway, where Codex runs
One agent covers the command line, the IDE, the web, iOS and GitHub Actions, and it reached Amazon Bedrock in June 2026 for enterprise stacks. Few developer tools stretch across that many surfaces at once. In practice, you kick off a task from your phone on the commute and pick up the result at your desk.
| Surface | What you do there |
|---|---|
| CLI (open source) | Pair programming in the terminal, local execution |
| VS Code extension | In-IDE edits, 9.8 million installs |
| Web | Background task delegation on your repos |
| iOS | Launching and monitoring tasks on the go |
| GitHub | Automatic pull request review |
Bundled with ChatGPT, metered in credits
Codex ships inside every ChatGPT plan, from Free to Pro. The $8 Go tier covers light work, Plus at around $20 opens the real toolbox (cloud delegation, GitHub, Slack), and heavy coders aim for Pro at $100 or $200. Teams go through Business, billed per seat.
Since April 2026, usage is measured in token-based credits that reset on a five-hour rolling window, so a small patch and a huge refactor no longer weigh the same. These figures shift with each OpenAI announcement, so treat the official pricing page as the only current source.
Frequently asked questions
Is Codex free?
Yes, a free ChatGPT account gives you trial access to Codex with no card required. The quota is small but enough to test the fit. Paid plans raise capacity and add the cloud features, such as background task delegation and automatic pull request review on GitHub.
Codex or Claude Code, which one wins?
Developer communities split by task rather than by tool. Threads on Reddit and Hacker News favor Claude Code for UI and architecture, and Codex for debugging, precise instructions and cost, with a token-efficiency edge often cited. Many engineers run both, and OpenAI even ships an official plugin that plugs Codex into Claude Code.
Does my code stay private with Codex?
It depends on the surface. The CLI works locally on your machine inside a sandbox, without uploading the repository. Delegated cloud tasks run in isolated containers where the repo is loaded for the duration of the job. Business and Enterprise plans add admin controls on top.
Can you run Codex with just an API key?
An API key covers the CLI, scripts and CI pipelines, billed per token through the OpenAI Platform. One caveat applies, some features tied to a ChatGPT account, including part of the cloud services, stay limited or unavailable under API-key authentication.
Verdict: Three agents chewing through a bug backlog in parallel is the kind of math that wins over busy engineering teams, and the free tier means you can verify the claim before spending a cent.
« An all-in-one creative platform that brings together more than 30 image and video models in a single interface. It includes camera controls that let you direct like a real filmmaker, characters that remain consistent from scene to scene, and ready-to-use tools for ads and short-form content »
