OpenAI updated Codeex this week, and the direction is unmistakable: the app is being positioned closer to a unified 'super app' or agent rather than a single-purpose coding tool. The new build folds in a collection of agentic capabilities that together feel like an operating layer on top of the Mac.
The headline feature is background computer use. Codeex can now spin up multiple agents that work on your Mac in parallel without disrupting your active work. Each agent gets its own isolated session, so you can keep writing, designing, or meeting in the foreground while Codeex refactors a codebase, scrapes a dataset, or runs a long QA pass in the background.
The app also ships with a built-in in-app browser, meaning agents no longer need to hijack your main Chrome or Safari session to browse, log in, or fill out forms. Combined with GPT image 1.5, now integrated natively, Codeex can generate, iterate, and drop images into any workflow without bouncing out to an external tool.
Perhaps the most interesting addition is comment-to-edit for websites. You can open a webpage inside Codeex, leave a comment on an element — 'make this heading bolder', 'move this CTA above the fold', 'swap this image for something warmer' — and the agent will directly edit the page. The interaction model feels closer to Figma comments than a chat prompt.
Taken together, the update is OpenAI's clearest signal yet that it views Codeex as the connective tissue between ChatGPT, GPT image 1.5, the new agent stack, and the user's desktop — a true super app rather than a productivity side project.