openai updated Codex Desktop to add a computer-use feature that can operate a user’s computer and an in-app browser. The additions push Codex beyond code generation into a productivity app that can run background automations, generate images and charts, and continue work over days or weeks. "We’re actually doing the sneaky thing where we’re building the superapp out in the open – and evolving it out of Codex." — Thibault Sottiaux (OpenAI’s Head of Codex).
Codex Desktop Computer-Use Feature
Codex Desktop now includes a computer-use feature that can run applications in the background and assign itself work, and the feature is available only for MacOS for now, which means Mac users can have Codex continue multi-step tasks without keeping every window active while other platforms are not yet supported.
Codex In-App Browser Feature
The app adds an in-app browser that lets a user click on an element and have the AI understand where the user is clicking, OpenAI did not demonstrate the browser performing automations during briefings, and Codex can attach automations to conversational threads so a user’s instructions and the resulting automated steps remain in the same chat flow.
Codex Memory and Automations
OpenAI said Codex can create an agent that generates an image, chart, or diagram automatically as part of an automation and can wake up to continue long-term work across days or weeks, and OpenAI described the app's memory as able to "remember useful context from previous experience, including personal preferences, corrections, and information that took time to gather" which supports those longer-running agent tasks.
OpenAI positions Codex Desktop as still targeted at programmers while expanding into broader productivity tools, and the company says 80% of its staff use Codex, indicating internal adoption even as the product broadens beyond its original coding-agent role.
When will the computer-use feature expand beyond MacOS and what pricing or access restrictions will govern its wider rollout?
