Anthropic has upgraded Claude's desktop computer use with a new background mode for Mac. The company is rolling out the feature in both Claude Cowork and Claude Code, giving Pro and Max customers a way to delegate real desktop tasks while continuing to use their Mac for something else.

Claude's computer use feature is not brand new. What changes with this update is the ability to run a computer task behind the scenes. Claude can click, type, open applications, move between windows, and finish a multistep assignment without making the user hand over the cursor. For people who have tried earlier versions of desktop AI agents, the new approach answers one of the biggest practical complaints: a tool that controls your mouse can only be used one task at a time.

Key facts

• Anthropic announced background computer use for Claude's Mac desktop app.
• The feature is available in Claude Cowork and Claude Code.
• It is limited to Pro and Max customers on macOS for now.
• Claude can control the mouse, type text, open apps, and manage workflows while users do other tasks.
• The update follows similar background computer use introduced earlier in the year for OpenAI's Mac assistant.

What is computer use in Claude?

Computer use is the name for Claude's ability to operate a graphical interface in the same way a person would. Instead of waiting for a programmer to build a custom integration, Claude looks at the screen, determines where controls are located, and uses synthetic mouse and keyboard events to perform actions. This gives the AI access to applications that have no public API and no dedicated plug-in. It also creates a more flexible agent that can navigate unfamiliar software.

That approach was always interesting for productivity, but the user experience was limited when Claude could only work in the foreground. The operator had to sit and watch, ready to intervene if Claude clicked the wrong setting or typed into the wrong field. It could still be useful for a single task, but it did not fit well into a busy workday. With background operation, Anthropic is trying to change the model from supervised helper to asynchronous coworker.

Available in Claude Code and Claude Cowork

Claude Code is Anthropic's command-line and development environment built for software engineering. Developers have used it to plan code changes, edit files, run tests, and execute shell commands. By giving Claude Code access to the entire desktop, Anthropic enables coding workflows that go beyond a repository. Claude can open a documentation site, copy a needed command, return to the editor, and keep moving through a task.

Claude Cowork is the broader desktop environment designed around working with Claude on documents, files, apps, and other everyday Mac jobs. The name suggests a collaborative workspace rather than a single AI chat window. With background computer use, Cowork becomes more useful for long-running business tasks such as organizing a folder, preparing a report, entering information into a web app, or processing a batch of images.

Both products share the same underlying model and permission system. The update means users can start an assignment in either environment, let Claude continue working, and turn their attention to a separate email, meeting, or creative project. On a laptop, this is especially valuable because screen space and user attention are limited.

Why background operation matters

The quality of an AI computer use feature is not only about whether the model can move a mouse. It is also about how the tool fits into a human workflow. Foreground agents are inherently serial. The user cannot type while the agent is typing. Background agents make parallel work possible. Claude can spend several minutes filling out a form or navigating an application while the user writes code, answers messages, or reviews a document in another space.

For many office tasks, this parallel structure is where the real efficiency gain appears. A person who delegates a repetitive job to Claude no longer needs to block out a portion of the day to supervise the AI. The Mac becomes a place where the user and the machine work on separate tracks. The economic value of AI assistants often depends on this kind of time shift, not merely on completing a single task a little faster.

Competition on the Mac

Anthropic is not entering a quiet field. OpenAI's Codex-turned-ChatGPT first brought background computer use to the Mac earlier in 2026. Since then, other AI companies have been experimenting with screen control, app orchestration, and autonomous desktop workflows. The Mac has become a test bed for agentic AI in part because macOS provides a stable environment where many apps can run at once and developers can offer deep system control in controlled ways.

The rise of assistant-driven computer use is being compared to the early days of smartphones. Instead of forcing every software company to build a different integration for each AI model, agents can operate the same apps people already know. This could lower the cost of software automation and put more pressure on traditional point-and-click workflows.

Who can access the update

Anthropic says background computer use is available for Pro and Max customers. This is a subscription-based rollout, which means users who are not on a Pro or Max plan will not see the feature. The company also notes that the capability is limited to macOS in this first release. There is no indication that Windows or other platforms are included in today's announcement.

Users who want to try the feature should confirm that Claude has the necessary desktop permissions and that the app is updated to the latest version. Because background operation can continue while the user steps away, it is also worth setting up clear tasks so Claude has a bounded job and does not have to make broad assumptions about what to do next.

Safety and supervision

Any technology that lets an AI agent press keys and click buttons on a real computer invites questions about safety, privacy, and accountability. When that agent runs in the background, users cannot catch every mistake in real time. That places a heavier burden on the model, the permission system, and the design of the user interface.

Anthropic has increasingly focused on agentic workflows, and computer use has been one of the most visible ways its models interact with real hardware. The decision to expand into background mode suggests confidence in the model's ability to follow instructions, recover from errors, and stay within boundaries. It also means users need to be thoughtful about the information they expose and the permissions they grant. Running a background agent with access to email, documents, and browser sessions is a significant act of trust.

What this means for the future of Mac AI

The new background computer use feature fits into a larger movement toward AI systems that are less like chatbots and more like employees. They do not simply answer questions; they take assignments, operate tools, and report back when finished. Claude's latest update is a signal that Anthropic believes desktop agents should be active participants in computing rather than passive text interfaces.

For Mac users, the practical benefits may become clear quickly. Tasks that once required screen recordings, keyboard shortcuts, or manual data entry can be passed to Claude. The AI can be asked to compare two documents, move files into specific directories, update contact records, or test a web page while the user does something else. These small forms of delegation can add up to meaningful time savings over a week.

The update also raises expectations for future releases. If background computer use proves reliable in Cowork and Code, similar features could appear across more Claude surfaces and perhaps on more operating systems. In the meantime, Anthropic has positioned Claude as a more useful companion on the Mac, one that can work in the same room without demanding everyone stop and watch.

Source: 9to5Mac News