OpenAI has released a preliminary research version of Codex, a cloud-based software development agent capable of performing multiple tasks simultaneously.
We’re launching a research preview of Codex: a cloud-based software engineering agent that can work on many tasks in parallel.
Rolling out to Pro, Enterprise, and Team users in ChatGPT starting today.https://t.co/HqiAtgydwh
— OpenAI (@OpenAI) May 16, 2025
The assistant is powered by codex-1, an optimized version of ChatGPT o3 for software development.
Initially, the agent is available to Pro, Enterprise, and Team users, with support for Plus and Edu promised “soon.”
“Codex can read and edit files, as well as execute commands, including test harnesses, linters, and type checkers. Task execution typically takes from a minute to half an hour depending on complexity,” the description states.
Each task is processed in a pre-loaded isolated environment.
Users can verify Codex’s work through quotes, terminal logs, and test results. In cases of uncertainty or failure, the agent reports issues for decision-making on further actions, OpenAI clarified.
“Like human developers, agents perform best when provided with configured environments, reliable testing setups, and clear documentation,” the team emphasized.
In April, OpenAI introduced the local software assistant Codex CLI.