Codex App: Computer Use Lets AI Agents Operate Mac Apps in the Background
CodexOpenAI introduced Computer Use in the Codex desktop app, enabling AI agents to see, click, and type across any macOS application using their own cursor — without interrupting the user's workflow. Multiple agents can run in parallel in the background, making it possible to have Codex test a frontend in a simulator, fill out a native GUI, or resolve a UI-only bug while the developer continues working. The feature is powered by a dedicated plugin (installed via Codex Settings) and is available on macOS at launch, with the European Economic Area, UK, and Switzerland excluded initially.
Key Takeaways
- Background operation is the headline differentiator: Codex runs agents with their own cursor without stealing focus, so the developer and the AI can work on the same machine simultaneously.
- Computer Use is plugin-activated: users must install the Computer Use plugin from Codex Settings before the feature becomes available — it is opt-in, not always-on.
- macOS-only at launch, with EEA, UK, and Switzerland excluded due to regional restrictions, limiting immediate reach for a significant portion of the developer population.
- OpenAI uses it internally to test Codex itself: the team has Codex verify its own implementations by running the resulting apps — a strong proof-of-concept for the reliability of the feature.
- Primary use cases are GUI-bound tasks: native app testing, simulator workflows, and bugs that only surface in native UI contexts where API access is impossible.
- Direct competitive pressure on Anthropic: TechCrunch framed the release as OpenAI "taking aim at Anthropic," whose Claude Code occupies similar developer mindshare — computer use extends Codex into territory Claude Code does not yet cover on desktop.
Sources & Mentions
5 external resources covering this update
OpenAI's Codex Mac app adds three key features that go beyond agentic coding
9to5Mac
OpenAI drastically updates Codex desktop app to use all other apps on your computer, generate images, preview webpages
VentureBeat
OpenAI takes aim at Anthropic with beefed-up Codex that gives it more power over your desktop
TechCrunch
OpenAI Codex Update Adds Computer Use, Image Generation, and Memory on Mac
MacRumors
OpenAI's latest Codex update builds the groundwork for its upcoming super app
Engadget
Computer Use Arrives in the Codex Desktop App
OpenAI shipped one of the most significant capability expansions in Codex's history with the April 16 release of version 26.415: Computer Use, a feature that gives AI agents the ability to operate any macOS application by seeing the screen, moving a cursor, clicking, and typing — all running in the background while the developer continues their own work.
What Computer Use Actually Does
Unlike browser-based automation tools, Codex's Computer Use integrates directly with the macOS desktop layer. Agents observe the screen visually, interpret what is displayed, and take targeted actions. OpenAI demonstrated the feature with a scenario where a user asks Codex to "check Slack, Gmail, Google Calendar, and Notion and tell me what needs my attention" — Codex navigates each app autonomously, gathers the relevant context, and reports back without the user switching focus once.
A key differentiator is the background operation model. Most computer-use implementations require exclusive focus on the machine, pausing or disrupting the user's own work. Codex uses an architecture that allows multiple agents to operate desktop apps in parallel while the developer continues coding or browsing in their own windows. As OpenAI's team put it internally, "It's literally Codex building itself" — the company uses the feature to have Codex verify its own implementations by running the resulting apps.
How to Enable It
Computer Use is available as a dedicated plugin. Users open Codex Settings, navigate to Computer Use, and click Install to activate the plugin. Once installed, prompting Codex to interact with a desktop app will trigger the computer-use capability automatically.
Use Cases
The primary use cases highlighted at launch include:
- Native app testing: automating simulator workflows and QA steps that would otherwise require manual repetition
- GUI-only bug resolution: addressing issues that only manifest in native UI contexts where API access is unavailable
- Cross-app information gathering: scanning multiple productivity apps (Slack, email, calendar) and synthesizing what requires attention
Availability and Regional Restrictions
Computer Use launched on macOS only. It is not yet available in the European Economic Area, the United Kingdom, or Switzerland at launch. OpenAI has indicated broader platform and regional availability is planned but has not confirmed a timeline.