OpenAI launches cloud Codex agent

OpenAI has introduced a research version of its cloud-based Codex agent — an AI-powered assistant capable of autonomously performing complex software development tasks. The tool is built on the codex-1 model, a version of ChatGPT o3 optimized specifically for coding workflows.

Autonomous Task Execution in an Isolated Environment

Codex can read and edit files, run commands, execute tests, check types, and more — all within a preloaded, isolated environment. Depending on complexity, tasks take between one minute and thirty minutes to complete.

The agent provides users with detailed outputs, including terminal logs, test results, and in-line code citations. If the system encounters a problem, it flags the issue for user review. Codex is currently available for Pro, Enterprise, and Team plan users, with support for Plus and Edu coming soon.