OpenAI announced a major Codex update on the 16th, moving the product to GPT-5.3-Codex and adding three headline features: background computer control, an in-app browser, and image generation. The release also includes 111 new plugins and a preview version of memory.
New features push Codex deeper into daily workflows
The most notable addition is Background Computer Use. According to OpenAI, Codex can operate Mac desktop apps in the background, handling actions such as opening Figma, clicking buttons, filling forms, and working across desktop software while the user continues doing other tasks. Multiple agents can run at the same time without interrupting one another.
The second major feature is In-App Browser, built on OpenAI’s Atlas technology. Codex can now act directly inside web apps, and users can leave annotations on a page to give precise instructions. Image generation has also been brought into the same environment. Powered by gpt-image-1.5, it can create product concept visuals, presentation assets, and prototype images without forcing users to switch tools.
Memory preview and plugin expansion arrive together
Alongside those features, OpenAI introduced a preview of memory for Codex. The system can retain useful information from prior work, including user preferences, corrections, and details that took time to gather, helping future tasks move faster.
OpenAI also expanded the plugin layer in the same release. The company said Codex is getting 111 new plugins, covering tools such as CodeRabbit, GitLab Issues, Slack, and Google Calendar. That broadens its reach across software development and workplace coordination.
Desktop execution becomes a larger part of the product
Taken together, the update shifts Codex beyond a text box and toward a more complete execution environment. The source article noted that Codex weekly active users grew 5x to 3 million over three months, a sign that demand exists for AI tools that can carry out work rather than only respond to prompts.
With desktop control, browser-based actions, image generation, and broader integrations bundled into one release, OpenAI is clearly putting weight on making AI agents part of everyday work routines.