Claude Can Now Control Your Computer

Anthropic launched a new feature today that lets Claude take direct control of your Mac to perform tasks on your behalf. It may be the most tangible agentic capability the company has delivered to everyday users so far.

The feature works by letting Claude click, type, and navigate apps as if a human were at the keyboard. Need to grab a file buried somewhere on your drive, or send a message while you're tied up with something else? Claude will find a way.

Anthropic is clear that the model always asks for permission before acting. You can stop it mid-task at any time. It can connect to services like Google Calendar and Slack via pre-built connectors, or use mouse and keyboard directly if no connector is available.

Security experts note that handing an AI access to your machine opens potential attack vectors. Anthropic says they've built in safety mechanisms, but recommends users think carefully about which tasks they delegate.

The feature is available to paying subscribers and marks Anthropic's response to the growing market for AI agents that can do things, not just answer questions.