Claude Now Controls Your Mac: Anthropic Introduces Computer Use Capability for Autonomous Task Completion
Key Takeaways
- ▸Claude can now directly control macOS computers, pointing and clicking to complete tasks autonomously with explicit user permission
- ▸Dispatch integration allows users to assign Claude tasks from mobile devices and retrieve completed work on desktop, enabling asynchronous task delegation
- ▸The feature includes safety mechanisms such as automatic detection of prompt injection attempts and restrictions on sensitive applications by default
Summary
Anthropic has announced a significant expansion of Claude's capabilities, enabling the AI assistant to directly control macOS computers and perform tasks autonomously. Through Claude's desktop and mobile applications, Claude can now use the mouse, keyboard, and screen to complete assignments—opening files, navigating browsers, and running developer tools without requiring users to set up integrations. The feature is rolling out in research preview to Claude Pro and Max subscribers, with built-in safeguards to detect prompt injection and prevent unauthorized access to sensitive applications.
The computer use capability pairs seamlessly with Dispatch, Anthropic's newly launched feature that allows users to assign Claude tasks from their phones while continuing work on desktop. This creates a continuous conversation experience across devices, enabling users to hand off assignments like morning briefings, code reviews, or data metric collection to Claude while away from their computers. While the capability is still early and subject to occasional errors, Anthropic is releasing it in preview to gather real-world feedback on performance and limitations.
- Computer use capability is currently in research preview for Claude Pro and Max subscribers on macOS only, with acknowledged limitations on complex task reliability
Editorial Opinion
Anthropic's introduction of computer use represents a meaningful leap toward AI agents that can interact with the digital world as humans do. By enabling Claude to manipulate screens and applications directly, the company is addressing a real gap in AI utility—many workflows still lack API integrations, and direct visual navigation offers practical value. However, the explicit focus on early-stage limitations and safety precautions suggests Anthropic is taking a measured approach to a genuinely powerful capability, which may be prudent given the broad implications of AI systems controlling user devices.
