OpenAI Expands Codex Into General-Purpose AI Workspace With Computer Use and 90+ Plugins

OpenAI is widening Codex far beyond its original role as a coding assistant. In a post on April 16, the company described "Codex for (almost) everything," an update that turns the product into a broader AI workspace with computer control on macOS, an in-app browser, scheduled automations, persistent memory, and support for more than 90 plugins.

Codex started as a tool aimed at developers who wanted help writing and editing code. The new version moves into general work automation, giving the agent access to everyday business software and the ability to act across multiple tools instead of staying inside an editor or chat window.

That shift matters because OpenAI is no longer treating Codex as just a programmer's sidekick. The company is positioning it as a work hub that can help with research, task handling, and cross-app actions, which puts it into direct competition with other agent products aimed at office work and IT teams.

The macOS computer use feature is one of the clearest signs of that change. It allows the agent to interact with a Mac more like a human user would, clicking, typing, and moving through interfaces when an API integration is not available or not enough for the task.

OpenAI also added an in-app browser, which gives Codex a built-in way to inspect websites and gather information without forcing every step through a separate browser session. That is a common pattern in agent systems, because many real-world workflows still involve a mix of web pages, dashboards, and internal tools.

The plugin list is also central to the update. OpenAI says Codex now supports more than 90 plugins, including integrations with Jira, Microsoft 365, Notion, and Slack. Those are the kinds of tools that sit at the center of product management, documentation, collaboration, and issue tracking in many companies.

Scheduled automations move Codex closer to a background worker than a chat assistant. Instead of waiting for a prompt, the system can be set up to run at defined times, which makes it more useful for recurring jobs such as gathering updates, checking tickets, or preparing summaries.

Persistent memory is another important piece. Rather than treating each interaction as isolated, Codex can retain context over time, which can reduce repetitive setup and make it more aware of ongoing projects, preferences, and previously completed work.

OpenAI's pitch here is not just that Codex can answer questions, but that it can participate in workflows. That places the product in a fast-growing category where vendors are trying to turn large language models into agents that can plan, remember, and act across software environments.

For security and operations teams, that kind of access is where the real questions begin. A system that can reach into calendars, tickets, documents, chat, and browser sessions is no longer only a productivity helper, it becomes part of the trust boundary around company data and actions.

OpenAI's plugin ecosystem and computer-use features suggest a product built for broader adoption rather than a niche developer audience. The company is making Codex look less like a code assistant and more like a general-purpose operator for digital work, with support for the tools and contexts that many teams already rely on.

That puts Codex in the same conversation as other agent platforms that combine memory, scheduling, and tool access, while also making it a more direct fit for everyday office workflows. The update's most notable detail is the breadth of integrations, which now includes more than 90 plugins across the collaboration and project-management tools many companies use every day.