OpenAI Expands Codex Beyond Coding With Computer Use, Memory, and New Workflow Tools

The Story

OpenAI has expanded Codex from a coding-focused tool into a broader software work assistant that can interact with desktop apps, access more parts of a computer, browse the web inside the app, and continue tasks over time. The update adds memory, more than 90 additional plugins, image generation, richer file previews, remote devbox access over SSH in alpha, and tools for tracking agent activity. OpenAI says more than 3 million developers use Codex every week. The rollout starts with Codex desktop app users signed in with ChatGPT, while some features will arrive later depending on platform, region, and account type.

The Facts

• OpenAI described this as a major Codex update aimed at making the product more useful across the full software development lifecycle.

• Codex can now go beyond coding and access other parts of a user’s computer.

  With background computer use, it can see, click, and type with its own cursor while operating desktop apps alongside the user’s own work instead of interrupting it.

• Codex now supports ongoing and repeatable work over longer periods.

  It can schedule future work for itself, wake up automatically to continue a task across days or weeks, and reuse existing conversation threads so prior context carries forward.

• A new memory feature is now in preview.

  Codex can retain useful context from earlier work, including preferences, corrections, and information that took time to gather. The feature is opt-in.

• Codex can use stored context to continue work more effectively.

  It can propose useful follow-up work based on project context, connected plugins, and remembered information, including surfacing open comments, relevant context, and prioritized next actions.

• Multiple agents can now work together on a Mac at the same time.

  This is intended to help with tasks such as frontend iteration, app testing, and workflows in software that does not expose an API.

• Codex now includes an in-app browser and is beginning to work natively with the web.

  Users can comment directly on pages to give precise instructions, which supports workflows such as frontend and game development. OpenAI says it plans to expand this beyond web applications on localhost over time.

• Codex can now generate images inside the same workflow.

  It uses gpt-image-1.5 to create and iterate on visuals that can be combined with screenshots and code for product concepts, frontend designs, mockups, and games.

• The update adds more tools for developer workflows.

  These include support for addressing GitHub review comments, reviewing pull requests, running multiple terminal tabs, and connecting to remote devboxes over SSH in alpha.

• Codex now supports richer file handling inside the sidebar.

  Users can open files directly with previews for PDFs, spreadsheets, slides, and documents, while a new summary pane shows agent plans, sources, and artifacts.

• OpenAI is also adding more integrations.

  Codex is getting more than 90 additional plugins spanning skills, app integrations, and MCP servers. Examples named include Atlassian Rovo, CircleCI, CodeRabbit, GitLab Issues, Microsoft Suite, Neon by Databricks, Remotion, Render, and Superpowers.

• Availability is rolling out in stages.

  The update is coming to Codex desktop app users signed in with ChatGPT. Personalization features such as memory and context-aware suggestions are coming soon for Enterprise, Edu, and users in the EU and UK. Computer use is launching first on macOS, with EU and UK rollout following, and no timeline was given for other operating systems.

Use Cases

Codex is being positioned as a broader software workflow tool rather than a code-only assistant. The update supports tasks such as frontend iteration, app testing, responding to GitHub review feedback, reviewing pull requests, creating mockups and product concepts, working across files and documents, connecting to remote development environments, and handling tasks that continue over time across tools such as Slack, Gmail, and Notion.

Timeline

This release adds several new capabilities at once: computer use, preview memory, recurring task scheduling, reuse of existing threads for automations, in-app browsing, image generation, richer file previews, a summary pane, pull request and GitHub review support, multiple terminal tabs, SSH connections to remote devboxes in alpha, and more than 90 new plugins. Not all features are available everywhere immediately, with some launching first on macOS or arriving later by region and plan.

Industry Context

The update also has a clear competitive dimension. As Anthropic’s Claude Code gains traction, OpenAI is expanding Codex across the same high-value workflow areas, suggesting a stronger push to compete more directly in agentic coding and developer productivity tools.

Why This Matters

This update pushes Codex from a coding assistant toward a broader software work agent that can act across apps, files, browser pages, remote environments, and longer task timelines. For end users, the shift is practical: less isolated code generation, more connected work completed inside one workflow. The update comes as OpenAI moves more aggressively to compete with the recent momentum of Claude Code.

This article was drafted with the assistance of generative AI. All facts and details were reviewed and confirmed by an editor prior to publication.