What OpenAI Codex does
Codex is OpenAI's coding agent for implementing changes, investigating bugs and reviewing code. You can work with it through a desktop experience, terminal, IDE integration or the web, depending on your account and setup. Local tasks use the project and tools available on your computer; eligible cloud workflows run in configured environments and can continue away from your laptop. Give it a specific outcome, repository context and verification steps, then review the resulting diff and checks. Codex is included across ChatGPT plans with different allowances and feature availability. Its open-source CLI and related components should not be confused with free, unlimited access to hosted models or cloud services.
Best for
- Developers delegating scoped implementation tasks
- Maintainers investigating bugs and reviewing pull requests
- Teams coordinating coding work across repositories
Key features
- Local and eligible cloud coding workflows
- Code implementation, debugging and pull-request review
- CLI, IDE and desktop entry points with tool integrations
Things to consider
- Free-plan access is limited and rollout-dependent; cloud access requires an eligible plan.
- API-key sessions have separate usage-based billing.
- Open-source client components do not make hosted inference or cloud usage unlimited.
Getting started
- Open an available Codex client and sign in with your ChatGPT account.
- Select a project or configured cloud environment and describe a bounded task.
- Review the changes and test results before applying them to your main branch.
