OpenAI Codex Goes Beyond Code: Full Mac Computer Use, Memory, and 90+ Plugins

OpenAI Codex Transforms Into a Full Desktop AI Agent

On April 16-17, 2026, OpenAI shipped what may be the most significant update to Codex since its rebirth as an agentic coding tool. The update, rolled out to macOS desktop users signed into ChatGPT, fundamentally changes what Codex is: no longer just a coding assistant, it is now a capable desktop automation agent able to see, click, and type across your entire Mac without any API integration.

Key Features

1. Computer Use on macOS

The headline feature is computer use. Codex can now operate Mac applications directly, using its own cursor to see what is on screen, click interface elements, and type into any app. Multiple agents can run in parallel without disrupting the user's own workflow. OpenAI positions this as a way to delegate complex, multi-step tasks that span several applications. EU and UK availability is planned for a later rollout phase.

2. Memory and Session Continuity

A new memory system allows Codex to retain context across sessions. It can remember individual user preferences, recurring workflows, tech stacks, and project-specific details gathered from previous conversations. The expanded automation system allows Codex to schedule future work and resume long-running tasks across days or weeks using existing conversation threads.

3. Built-In Browser

Codex now ships with an integrated browser. Users can navigate to any page, comment directly on elements, and provide precise instructions to the agent. OpenAI notes the browser is especially useful for localhost frontend work and game development, where visual context is critical.

4. Image Generation via gpt-image-1.5

Codex integrates OpenAI's latest image model, gpt-image-1.5, directly into the development workflow. Developers can generate and refine images alongside code, combining screenshots and source files to create mockups, game assets, and product visuals without leaving the environment.

5. 90+ New Plugins

The plugin ecosystem expanded dramatically, adding integrations with Atlassian Rovo for Jira, CircleCI, CodeRabbit, GitLab Issues, and the full Microsoft Suite. The new plugin architecture combines skills, app integrations, and MCP servers to improve context gathering and action execution.

Usability Analysis

For developers working on macOS, this update represents a substantial shift in how an AI coding tool fits into the daily workflow. Tasks that previously required careful API setup or manual hand-off between tools can now be orchestrated through Codex directly. The memory system is arguably the most practically impactful addition: the ability to carry context across sessions eliminates the repetitive onboarding that has frustrated long-term users of AI coding assistants.

The computer use feature, while powerful, comes with important caveats. It is macOS-only at launch, EU and UK users are excluded, and enterprise and education deployments require a separate rollout. This limits its immediate applicability for teams in regulated regions. The parallel agents capability addresses a long-standing productivity concern: AI tasks no longer block the user's own work on the machine.

Pros and Cons

Pros:

• Computer use without API integration lowers the technical barrier for automation
• Persistent memory eliminates repetitive context-setting across sessions
• 90+ new plugins cover the most common developer tool stacks
• Parallel agent execution does not block user workflow
• gpt-image-1.5 integration brings visual design into the coding loop

Cons:

• Computer use is macOS-only; Windows users are not yet supported
• EU and UK markets excluded at launch due to regulatory considerations
• SuperGrok-style premium tier dependency is unclear; pricing was not disclosed
• Memory is described as a "preview feature," suggesting limited reliability at launch

Outlook

The Codex update signals OpenAI's intent to position Codex as the primary AI operating layer on the developer's machine rather than a specialized coding tool. By combining computer use, memory, and a rich plugin ecosystem, OpenAI is competing directly with emerging desktop AI agent frameworks. The natural next step is a Windows release and full EU/UK availability, which would dramatically expand the addressable user base. Integration with GPT-6, currently in pre-training, could further elevate agent capabilities once that model becomes generally available.

Conclusion

OpenAI's April 2026 Codex update is a genuine expansion of scope. Developers on macOS now have access to an AI agent that can operate their entire desktop, remember their workflows, generate images, and integrate with the tools they already use. The EU/UK limitations and macOS exclusivity are notable gaps, but for users in supported regions, this update materially changes the value proposition of subscribing to Codex.

Editor's Verdict

OpenAI Codex Goes Beyond Code: Full Mac Computer Use, Memory, and 90+ Plugins earns a solid recommendation within the gpt space.

The strongest case for paying attention is full Mac computer use without any API integration required, which raises the bar for what readers should now expect from peers in this space. Reinforcing that, persistent memory eliminates repetitive context-setting adds practical value rather than just headline appeal. The broader signal worth registering is straightforward: openAI is repositioning Codex from a coding assistant to a full desktop AI agent, competing with emerging agentic frameworks. On the other side of the ledger, macOS-only at launch; no Windows support announced is a real constraint, not a marketing footnote, and it should factor into any serious decision. Layered on top of that, EU and UK excluded due to regulatory considerations narrows the set of teams for whom this is an obvious yes.

For ChatGPT power users, OpenAI API customers, and enterprise teams already running on the OpenAI stack, this is a serious evaluation candidate, not just a curiosity to bookmark. For everyone else, the safer posture is to monitor coverage and revisit once the use cases that matter to your team are demonstrated in the wild.