Anthropic’s Claude Code and Cowork can control your computer
Why it matters: This marks a leap towards truly autonomous AI agents, blurring lines between human and machine interaction.
- Anthropic’s Claude Code and Cowork can now autonomously control macOS devices to perform tasks, leveraging the computer's browser, mouse, and keyboard when direct integrations aren't available.
- The new feature is a research preview limited to macOS and available for Claude Pro and Max subscribers, requiring the desktop app to be running and paired with the mobile app.
- Claude will "always ask for your explicit permission" before exploring, scrolling, and clicking, and its "auto mode" is designed to prevent destructive actions like mass file deletion, according to ZDNET.
- Anthropic acknowledges that complex tasks may require retries and screen-based execution is slower than direct integrations, sharing the feature early to gather feedback on its performance and limitations.
Anthropic has significantly upgraded its Claude Code and Cowork AI tools, enabling them to autonomously control macOS computers to complete tasks like opening files, browsing, and running dev tools, even when users are away. This "auto mode" builds on previous autonomous capabilities, allowing the AI to make permission-level decisions while preventing destructive actions, though it's currently a research preview for Pro and Max subscribers.
