Take on real engineering work
Give Codex difficult refactors, hidden defects and new feature work, and get high-quality results with fewer tokens.
Codex is OpenAI’s AI coding agent. Describe the task and it can implement features, refactor complicated code, fix hard-to-find bugs, run checks, and review its own work.
Downloads, system requirements, plans and the latest features are governed by OpenAI’s official page.
Task
Refactor the payment module and add the missing tests.
Capabilities
From hard-to-find defects to cross-file refactors, Codex does more than generate snippets: it moves a complete engineering task forward and verifies the result.
Give Codex difficult refactors, hidden defects and new feature work, and get high-quality results with fewer tokens.
Run multiple agents in parallel, let them use computer and browser tools to verify the work, then track and review everything in one place.
Close your laptop and tasks can keep running. Each repository gets a reusable cloud environment, so you can pick up from another device.
Codex can review pull requests automatically, point out issues and answer questions, leaving judgment calls to your team.
Surfaces
Move between ChatGPT, your editor and the terminal while your ChatGPT account keeps tasks and context connected.
Start tasks, follow progress and review results from one workspace.
Keep tasks running in the cloud and resume from a web page or another device.
Bring code suggestions, task execution and change review into your editor.
Run Codex in your terminal and bring the agent into your existing development flow.
npm i -g @openai/codexWith Coskey
If you want Codex to use your own model providers, Coskey manages providers, model mappings and session routing without modifying the app.
Official source
This site only introduces Codex and does not host installers. Visit OpenAI for the latest versions, supported platforms, plans and complete feature details.