OpenAI updates Codex with computer use, a built-in browser, memory and plugins

OpenAI has released a major update to the Codex desktop app for macOS and Windows. The new version adds five key features: computer use, a built-in browser for working inside the app, image generation, long-term memory across sessions and plugin support. The goal is to speed up development workflows and turn Codex from a smart autocomplete into a full agent that carries out tasks on its own.

Computer use is the most significant addition. Codex can now interact directly with the operating system: run terminal commands, manage the file system, open other applications and execute complex sequences of actions without the developer's involvement. This is a fundamental shift: Codex used to only advise, and now it acts. A developer only needs to state a task at a high level, for example "set up CI/CD for the new project," and the agent works out the steps and carries them out itself. This matters especially for DevOps tasks, which previously required a separate set of tools.

The built-in browser addresses one of development's chronic pain points. Programmers constantly switch between the editor and the browser: looking up current documentation for a library, checking ready-made solutions on Stack Overflow, reading the changelog of a new framework version. Codex takes this search on itself: the agent finds the information, analyzes it in the context of the current code and applies the result right away, without interrupting the workflow. This is particularly valuable when working with APIs, where backward compatibility often breaks between versions.

Image generation extends Codex's audience beyond backend development. You can now create screen mockups, icons or placeholder images right in the working environment, without switching to Figma or a separate image generator. For frontend teams, this removes the constant context gap between code and design. In early prototyping, where speed matters more than final quality, this integration saves time and cuts out extra hops between tools.

The memory feature changes the nature of working with the tool. Codex now remembers architectural decisions made in a project, code style conventions, details of the working environment and the context of previous tasks. The longer a developer works with the app, the more precise the agent's suggestions and actions become. That is a fundamental difference from stateless tools, which start from scratch every time. Plugins complete the picture: developers will be able to connect corporate systems, in-house tools and external services to Codex, tailoring the workflow to a specific team's needs. Each connection extends the agent's capabilities without requiring changes to the core app.

The update clearly signals OpenAI's strategic direction: from a language model to an agentic working environment. Codex now competes not only with GitHub Copilot but also with a new generation of tools: Cursor, Windsurf and Devin. The company is steadily moving the product from the "assistant" category to the "autonomous executor" category. Developers should watch this transformation: tools that look cutting-edge today may become the standard in a year. Those who master agentic workflows now will gain a real advantage.