Anthropic’s Claude Code and Cowork can control your computer

Anthropic has enhanced its Claude AI with new autonomous capabilities in its Code and Cowork tools, allowing the AI to control a user's computer to perform tasks. This feature, currently available as a research preview for Claude Pro and Max subscribers, is limited to macOS devices for now. The system works by prioritizing connections to supported services like Slack and Google Workspace, but if connectors are unavailable, Claude can directly control the browser, mouse, keyboard, and display to execute tasks. Anthropic assures users that Claude will always request explicit permission before interacting with the screen. This functionality builds upon autonomous features introduced with the Claude 3.5 Sonnet model and works particularly well with the Dispatch feature, though the company acknowledges that complex tasks might require retries due to the slower nature of screen interaction compared to direct integrations.

(Source：The Verge)