Anthropic has rolled out a new feature for its Claude AI assistant that takes automation a step further—allowing the system to directly operate a user’s computer with minimal input.
According to Felix Rieseberg, who leads engineering for Claude Cowork and Claude Code Desktop, this capability enables Claude to interact with a computer much like a human would. It can use the mouse, type on the keyboard, and read what’s on the screen, meaning it can handle tasks across virtually any application.
The feature becomes even more powerful when used alongside Dispatch. This tool lets users assign tasks remotely, so Claude can continue working on their computer even when they’re not physically present.
In practical terms, Claude can open apps, browse the web, and update spreadsheets—essentially carrying out everyday desk tasks on behalf of the user. For now, this functionality is available as a research preview within Claude Cowork and Claude Code, specifically for Pro and Max users on macOS.
Anthropic has also built in safeguards. Claude is designed to first rely on integrations with connected tools like Slack and Calendar. If a direct integration isn’t available, it will ask for user permission before accessing and controlling an app on the screen.
Another key aspect of the update is its ability to handle tasks asynchronously. Users can assign work from one device, such as a smartphone, and return later to find it completed on their computer. This reflects a broader shift toward AI systems that can take on delegated tasks rather than simply respond to prompts.
The update builds on Anthropic’s earlier launch of its Cowork platform, which aims to function as a digital coworker capable of managing workflows across files and applications. With this move, Claude continues to expand beyond coding support into more general productivity use cases, including document handling and report creation.
Also Read: Anthropic Launches Feature to Transfer Memory From ChatGPT and Gemini to Claude
