Claude AI Can Now Operate a Mac Without Touching It

Claude comes with a more closed and controlled system. (photo: x @AMohameedsardar)

JAKARTA - Anthropic is shaking the AI world again with its latest update. Through the new feature in the Claude AI, now this smart assistant not only answers questions - but can also directly operate the user's Mac computer, even when left behind.

This feature is available through Claude Code and Claude Cowork, which allows AI to act like a human on the screen. When it doesn't have direct access to a specific application, Claude will "see" the screen, then click, open files, and run the browser automatically.

In other words, this is not just AI - it has entered the level of "real digital assistant" that can complete work without direct user intervention.

New Claude AI Features

This ability is integrated with the Dispatch feature, which allows users to give tasks from the iPhone, then return to the Mac with the work completed. In an official demonstration, Claude was even able to export a presentation into a PDF and immediately send it to a meeting invitation - without being touched at all.

However, Anthropic admitted that this technology is still in its early stages. This "computer use" system is not as accurate as Claude's ability to code or text, and still has the potential to make mistakes.

Therefore, users are advised not to immediately entrust sensitive data. The security approach used is also permission-based, where Claude must ask for access before interacting with new applications.

This feature is often compared to OpenClaw, an open-source AI that went viral because of its ability to control devices and applications such as WhatsApp or Telegram. But unlike OpenClaw which is flexible but complicated, Claude comes with a more closed and controlled system.

Currently, the feature is only available in the research preview stage for Claude Pro and Max customers, and is still limited to Mac devices.