OpenAI Codex 2.0: dedicated autonomous coding agent in ChatGPT and API

Think of Codex 2.0 as a tireless programming assistant you can assign a whole task to and walk away from. You tell it "fix the bug in the authentication module, add tests, and open a pull request" — and it actually does that, on its own, start to finish.

OpenAI built this not as a code autocomplete tool, but as a true agent: it takes on complete coding assignments across an entire project. It works inside a virtual machine running in OpenAI's cloud — a sandboxed environment where it can safely do its work without touching your own computer or server.

The agent reads through your project files, understands what the code is supposed to do, writes the necessary changes, runs the test suite to check nothing is broken, and then opens a pull request for a human to review before anything gets merged. It is not a chatbot that suggests lines of code — it is closer to a contractor who takes a ticket and delivers finished work.

What sets it apart from similar tools like Anthropic's Claude Code is the delivery model: Codex 2.0 runs entirely in the cloud, with no local installation needed. You activate it from ChatGPT or via an API call, and it operates remotely. This makes it especially appealing for development teams and companies looking to automate chunks of their workflow without changing their local setup or dev environment.