Claude from Anthropic has mastered controlling a Mac computer on behalf of the user

Anthropic expands Claude’s capabilities: it can now control a Mac computer

Anthropic has introduced a new feature for its Claude model—direct interaction with the macOS desktop. In Cowork and Code modes, the AI can open files, launch browsers, use development tools, and perform any actions that a user would normally do.

How It Works
* Autonomous Computer Control

Claude can independently open the required document, edit it, or carry out more complex operations without human intervention.

* Task Transfer from Mobile Devices

Through the Dispatch feature, commands can be sent to a Mac from a smartphone. This allows control of the model’s work even when you’re elsewhere.

Demonstration
Anthropic posted a YouTube video where Claude showcases its skills: opening and editing files, launching a web browser, interacting with an IDE. All of this happens “as if” the user is sitting at the computer.

Availability
* The feature will be available on macOS only for paid Claude Pro and Claude Max subscriptions.

* Currently it is in research preview access, meaning it can only be used by a limited group of users.

Why It Matters
Anthropic has long been developing AI agents. Now the company itself implements technology that was previously associated with OpenAI (OpenClaw). Other players have entered the market—Perplexity, Computer, and even Meta through the Manus project.

Thus, Claude gains a new level of autonomy: it not only answers questions but also controls real devices, opening possibilities for deeper user interaction.