OpenAI has launched ‘Codex’, a cloud-based AI agent designed to assist with software engineering tasks such as writing code, fixing bugs, answering questions about codebases, and suggesting pull requests.

Codex is powered by Codex-1, a version of OpenAI’s o3 model specifically designed for software development. Trained in reinforcement learning on real-time coding tasks, Codex can write code that mimics human developers, accurately follow instructions, and run tests until the code passes. The AI ​​tool is now available for ChatGPT Pro, Enterprise, and Team users with Plus and Edu support.

Each Codex task runs in a secure, isolated environment preloaded with the user’s codebase. The agent can read and edit files, run test suites, linters, and type checkers, and tasks typically complete in 1 to 30 minutes, depending on complexity. Users can also monitor its progress in real time.