Anthropic Turns Claude Into an AI Agent That Can Control Your Computer

Claude Can Now Operate Your Computer

Anthropic has rolled out a major upgrade to its AI assistant Claude, transforming it into a fully functional agent capable of interacting directly with a user's computer. The bot can launch applications, navigate web browsers, fill in spreadsheets, and handle a wide range of everyday desktop tasks.

«You can now enable Claude to use your computer to complete tasks. It opens your apps, navigates your browser, fills in spreadsheets — anything you'd do sitting at your desk. Research preview in Claude Cowork and Claude Code, macOS only.» — Claude (@claudeai), original post

Users can issue commands to the digital assistant from their phone, and Claude will carry them out on the desktop. As a demonstration, Anthropic showcased a scenario where a user was running late for a meeting. They asked Claude to export a presentation to PDF and attach it to a calendar invite — the bot completed the task successfully.

Why This Matters

The transformation of Claude into an AI agent that directly interfaces with the operating system marks a pivotal shift from text-based chatbots to autonomous digital assistants. Agents like this have the potential to fundamentally reshape workflows — from automating repetitive operations to executing complex multi-step tasks with minimal human oversight. The launch intensifies the already fierce competition between Anthropic and OpenAI for dominance in the AI assistant market.

How the Agent Mode Works

According to Anthropic's website, users can send a prompt from their smartphone, step away to handle other things, and return to find the work completed on their computer. Claude can be instructed once to check email every morning or compile a report on Fridays, and it will handle the task on a recurring basis.

Anthropic has implemented multiple layers of safety:

• Claude always requests permission before accessing new applications;
• Before executing any task, the bot presents a plan and waits for user approval;
• At any stage, users can refine the task or choose a different approach.

Limitations and Caveats

Anthropic acknowledged that computer-use functionality remains at an early stage compared to Claude's more mature capabilities in coding and text processing. Company representatives warned that the bot can make mistakes, and that security threats continue to evolve despite ongoing improvements to safety measures.

Anthropic stated that the feature was built with "risk minimization" in mind. The service is available through Claude Cowork and Claude Code, currently limited to macOS. Users need to navigate to the Cowork section to activate it.

Claude's Market Position

The agent mode launch comes amid Claude's surging popularity. In March, the application climbed to the number one spot among free apps in the US App Store, pushing OpenAI's ChatGPT down to second place.