OpenAI's latest major Codex update brings computer use to better compete with Claude Code

OpenAI has released a major update to Codex, significantly expanding its abilities across the entire software development lifecycle. Codex can now operate computers by controlling all installed applications, using its own cursor to see, click, and type. Multiple Codex agents are able to run in parallel on Mac devices, working alongside users without interfering with other tasks.

Alongside these system-level changes, Codex introduces deeper integrations for developer workflows. The app now supports pull request reviews, viewing multiple files and terminals, connecting to remote development environments over SSH, and includes an in-app browser for faster iteration on frontend designs. This browser also allows direct commenting on web pages to provide detailed instructions to Codex agents. Users can address GitHub review comments, open files with rich previews, including PDF, spreadsheet, slide, and document formats, and monitor agent activity through a summary pane.

Building on these improvements, Codex now supports GPT-Image-1.5 for generating and refining images. More than 90 new plugins further extend its integration with apps and MCP servers. Automations now support persistent conversation threads, and a new memory preview enables Codex to remember context from previous sessions. These features are currently rolling out to desktop app users with ChatGPT accounts, with broader personalization and computer use for Enterprise, Education, European Union, and United Kingdom users coming soon.