Anthropic Launches Computer Use: Claude Now Operates Desktops Autonomously

A New Era of AI Agency

Anthropic has launched a research preview that gives its Claude AI the ability to see and operate computers much like a human does. By using the new Claude Cowork and Claude Code desktop applications, the AI can now move cursors, click buttons, and type text to complete complex workflows. This capability represents a significant shift from simple chat interfaces to active agentic assistance. It allows the model to bridge the gap between thinking about a task and actually executing it in a local environment.

Mobile Control via Dispatch

The feature integrates with a new mobile tool called Dispatch, which allows users to assign tasks from their phones for the desktop to execute. Whether it is summarizing emails during a morning commute or running code tests while away from the desk, Claude functions as a persistent background worker. This remote hand-off ensures that productivity continues even when a user is not physically present at their machine. Users simply pair their devices via a QR code to start delegating work.

Prioritizing Security and Precision

Claude prioritizes using direct API connectors for speed but falls back to visual screen navigation when those are unavailable. To maintain security, Anthropic includes automatic scans for prompt injections and requires explicit permission before the AI accesses new applications. Users also retain a universal stop button at any moment if they need to intervene. Certain sensitive applications are restricted by default to minimize potential risks during the research preview.

Empowering Technical Workflows

For software engineers, Claude Code offers deep integration with IDEs and terminal commands to manage pull requests or fix bugs. It can autonomously read through codebases and run local test suites to ensure stability before a human reviews the final output. This reduces the manual toil often associated with repetitive development cycles and complex environment setups. By automating these lower-level tasks, developers can focus on high-level architecture and creative problem-solving.

Future Outlook and Requirements

Currently available for Claude Pro and Max subscribers on macOS, the technology is still in an early research phase. Anthropic expects future versions to become faster and more reliable as the model learns to navigate increasingly complex user interfaces. While complex tasks may occasionally require a second try, the current rollout sets the stage for a world where AI acts as a dedicated local task runner for every professional. We can expect expanded support for more operating systems and deeper application integrations in the coming months.