OpenAI is widening the role of Codex from an AI coding assistant into something closer to a software workbench. In a new update, the company says Codex can now operate apps on a user’s computer, work with more external tools, generate images, remember preferences, and take on repeatable work over time.

That framing matters because it shifts Codex from a product focused mainly on writing code toward one aimed at the full software development lifecycle. OpenAI is betting that developer productivity gains will come less from isolated code generation and more from reducing the friction between coding, reviewing, testing, design iteration, and project follow-through.

Codex Is Moving Into the Rest of the Workflow

The biggest change is that Codex can now act more directly across the developer environment. OpenAI says the app supports background computer use on macOS, allowing agents to see, click, and type with their own cursor while multiple agents run in parallel without interrupting the user’s own work in other apps.

That is a meaningful product expansion. Many bottlenecks in software work do not live inside the code editor alone. They show up when a developer needs to inspect a UI, test a flow in another app, coordinate across browser tabs, or work through tools that were never designed with clean APIs in mind. OpenAI is clearly trying to make Codex useful in those spaces rather than leaving them outside the product boundary.

The app is also getting an in-app browser, which lets users comment directly on a page to guide the agent more precisely. For frontend work especially, that creates a tighter loop between seeing a problem and instructing the model to fix it.

OpenAI Wants Codex to Handle More Than Code

The release also adds support for gpt-image-1.5, which means Codex can generate and iterate on visuals inside the same workflow developers already use for apps, product concepts, mockups, and games. That is not a side feature. It reflects a broader view of modern development, where product work often spans interface design, screenshots, assets, and rapid experimentation alongside code.

OpenAI is pairing that expansion with more than 90 new plugins that combine skills, app integrations, and MCP servers. The company specifically highlights developer-facing integrations such as Atlassian Rovo, CircleCI, CodeRabbit, GitLab Issues, Microsoft Suite, Neon by Databricks, Remotion, Render, and Superpowers. In other words, the update is not just about a smarter model. It is about putting Codex closer to the actual toolchain where software work happens.

The Deeper Bet Is on Continuity

Some of the most important additions are about persistence rather than moment-to-moment assistance. OpenAI says Codex can now reuse existing conversation threads, preserve built-up context, schedule future work, and wake itself up later to continue longer-running tasks. It is also previewing memory, so the system can retain preferences, corrections, and context gathered from previous work.

That could be more important than any single workflow feature. AI developer tools become more valuable when they do not have to be re-taught the same context every day. If Codex can carry state across sessions and proactively suggest the next useful action, it starts to look less like a chatbot for coding and more like an ongoing collaborator inside a team’s development rhythm.

OpenAI also says the app now supports richer review and coordination workflows, including GitHub review comment handling, multiple terminal tabs, SSH access to remote devboxes in alpha, file previews for documents and spreadsheets, and a summary pane for plans, sources, and artifacts. Taken together, those changes suggest OpenAI wants Codex to feel more like a persistent operating layer for technical work rather than a single-purpose assistant.

Why This Release Matters

The larger story here is product convergence. Coding assistants, browser agents, design helpers, and workflow automation tools are starting to collapse into one category. OpenAI is trying to position Codex at the center of that shift by making it capable across code, interfaces, assets, tools, and follow-up work instead of limiting it to prompt-in, code-out interactions.

For now, the rollout starts with signed-in ChatGPT users on the Codex desktop app, with some personalization features and regional availability expanding later. But the strategic direction is already visible. OpenAI wants Codex to be useful not only when a developer asks for code, but across everything that surrounds building software day after day.

You can also view OpenAI’s launch post on X here.

Developers who want to explore the updated experience can start from the main Codex product page.

Comments

No comments yet. Be the first to share your thoughts.

or to leave a comment.