OpenAI Codex Adds Screen Control and 90+ Plugins, Becoming an AI Workspace

OpenAI's Codex has always been positioned as a developer tool — write code, fix bugs, review pull requests. But its latest update, detailed in a blog post titled "Codex for almost everything," signals a deliberate pivot: Codex is no longer just for coding. The tool now controls your screen, integrates with 90+ business apps, schedules its own work, and can even watch what you're doing to build contextual memory, according to XDA Developers.

"It's clear from the way they're talking about Codex that OpenAI doesn't see it as just a coding tool anymore," wrote Mahnoor Faisal at.¹ The blog post title alone tells the story: Codex for almost everything.

The headline feature is background computer use on macOS. Codex can now operate graphical interfaces independently — clicking buttons, typing into fields, navigating apps — using its own cursor that runs separately from yours. This isn't just for testing frontend changes or working with tools that lack APIs, which is how OpenAI frames it. As XDA notes, if Codex can navigate any app on your Mac, there's nothing stopping it from managing spreadsheets, organizing files, or handling admin work that has nothing to do with code.

The feature is currently unavailable in the United Kingdom, Switzerland, and the European Economic Area, according to XDA. Sam Altman explicitly encouraged people on X to try computer use for non‑coding tasks — things like skipping to the next song on Spotify or migrating data from a document to a spreadsheet. Multiple agents can run in parallel, each handling a different task. The feature is currently macOS‑only and unavailable in the UK, Switzerland, and the European Economic Area.

Alongside screen control, OpenAI released over 90 new plugins covering tools like Jira, GitLab, CircleCI, and the Microsoft Suite. But the more interesting update is Codex's ability to schedule its own work. Per,¹ Codex can now set future tasks for itself, wake up automatically, and continue working across days or weeks. It can also proactively suggest what you should work on next by pulling context from your codebase, Google Docs, Slack, Notion, and any connected plugins.

"So instead of you opening Codex and telling it what to do, Codex can open up and tell you what needs doing," ¹ observed. That's a significant shift — from reactive tool to proactive assistant.

The most intriguing addition is Chronicle, currently in opt‑in research preview for ChatGPT Pro subscribers on macOS. Chronicle watches your screen in the background, takes periodic screenshots, and converts what it sees into structured memories stored as local Markdown files. The screenshots are deleted after six hours, but the memories persist — meaning the next time you open Codex, it already knows what you've been working on.

OpenAI says Chronicle is designed to reduce the amount of context you have to restate. Sandboxed agents process the screen captures in the background to generate the memories. The privacy implications are significant — a tool that watches everything on your screen raises obvious questions — but the capability signals where OpenAI is heading: an AI that understands your workflow rather than just executing tasks.

Codex's expansion beyond coding lands in the middle of a heated competition. Anthropic's Claude Code has been the darling of the developer community, but it's facing its own turbulence — Anthropic has been quietly testing whether to remove Claude Code from its $20 Pro plan, as reported by XDA. Meanwhile, OpenAI is pushing Codex in the opposite direction: more features, broader scope, and free tier access.

The competitive dynamic is striking. Anthropic is restricting access to manage infrastructure strain. OpenAI is expanding access, using free‑tier Codex as a funnel. For builders, this creates an unusual moment: the tool with the most generous free tier is also the one adding the most ambitious features.

First, screen control changes the definition of an AI coding tool. When an AI can navigate any app on your machine, the line between coding assistant and general‑purpose AI agent disappears. Second, self‑scheduling agents mean AI that works while you sleep. Codex can now run tasks across days or weeks without you touching it. Third, Chronicle‑style memory is the next frontier. The AI that remembers your context without being told is fundamentally different from one that starts fresh every session.

OpenAI's closing line in the Codex blog post talks about "narrowing the gap between what people can imagine and what they can build," as OpenAI wrote in its.¹ But the features they're shipping suggest a more specific vision: an AI workspace that happens to do coding, rather than a coding tool that happens to do other things. For builders, this blurring of boundaries is both opportunity and risk. More capability in one tool means less context‑switching. But it also means deeper integration into a single provider's ecosystem — and all the lock‑in that implies.