OpenAI has released Codex, a research preview of its latest coding agent, which is now integrated into ChatGPT. The Codex agent runs in a sandboxed, virtual computer in the cloud. By connecting with GitHub, Codex’s environment can come preloaded with your code repositories. OpenAI says the AI coding agent will take anywhere from one to 30 minutes to write simple features, fix bugs, answer questions about your codebase, and run tests, among other tasks.

Codex is built on codex-1, a fine-tuned version of OpenAI’s o3 model. According to OpenAI, this model produces more accurate and instruction-following code compared to earlier iterations and can repeatedly test its code until it reaches a passing output. Under the ChatGPT web app, users can access Codex from the sidebar. Two interaction modes are offered: “Code” for generating code and “Ask” for asking questions. Codex keeps a visible log of all tasks and their progress within the same panel.

At first, OpenAI is offering broad access to Codex at no additional cost for current eligible users. However, the company has stated that rate limits will be introduced in the coming weeks, with additional usage available through a credit-based system. Codex does not connect to the internet or external APIs during operation. All tasks are performed in an isolated environment to prevent unintended access to online resources or services.

OpenAI has confirmed that it will expand access to ChatGPT Plus and Edu users in the near future, though no specific timeline has been announced. This release follows recent developments from other major AI companies. For example, Anthropic launched its Claude Code tool in February, and Google updated Gemini Code Assist in April with additional capabilities aimed at developers. OpenAI’s launch of Codex adds to this growing field of tools designed to support software engineering through automation and natural language guidance.