Anthropic has updated Claude Code and Claude Cowork with computer-use capabilities for macOS, allowing Claude to autonomously control your browser, mouse, and keyboard to complete tasks like checking emails, compiling spreadsheets, or opening pull requests. Claude prioritizes direct service connectors (Slack, Gmail) before falling back to screen interaction. Security guardrails include per-app permission prompts, a blocklist, prompt injection scanning, and training to avoid risky actions like fund transfers or file deletion. However, Anthropic acknowledges risks remain — Claude takes screenshots of whatever is on screen, so users should close sensitive files before use. The feature is in research preview for Claude Pro and Max subscribers on macOS only, with gradual rollout planned as Anthropic gathers real-world feedback.
Sort: