Anthropic Enables Claude to Control macOS Desktop as Research Preview

Anthropichas introduced a new capability for Claude: the ability to directly control a user's macOS desktop to complete tasks. The feature is now available as a research preview in Claude Cowork and Claude Code, marking a significant expansion of Claude's operational scope beyond text generation.

How It Works

Claude's desktop control operates with a fallback strategy. According to Anthropic, the model first attempts to use existing integrations with connected apps like Slack, calendars, and other services. It only assumes direct control of the desktop when no standard integration is available, making desktop operation a last resort rather than the default behavior.

Capabilities include opening applications, navigating web browsers, filling spreadsheets, and performing other manual desktop tasks. The feature currently supports macOS only.

Anthropichas also rolled out "Dispatch," a companion feature that enables users to remotely control their own computers from any location.

Built on Vercept AI Acquisition

The desktop control technology origins from Vercept AI, a startup Anthropic acquired approximately four weeks before this announcement. According to Vercept co-founder Kiana Ehsani, the team shipped its first product in less than four weeks post-acquisition—faster than anticipated.

Ehsani credited Anthropic's organizational culture for the rapid development: "Everyone moves fast, everyone is incredibly smart, humble and supportive, and it's really easy to get things done," she stated on X (formerly Twitter). She identified Anthropic's talent base as its primary competitive advantage.

Unresolved Security and Reliability Questions

The feature remains in research preview status, with substantial questions unresolved regarding real-world reliability and security. Granting AI full desktop access creates multiple risk vectors: data privacy exposure, potential error cascades, system control failures, and expanded attack surfaces.

Anthropicwill need to demonstrate robust safeguards before broader deployment. The company's more ambitious approach to full desktop control contrasts with earlier attempts at browser-only automation. OpenAI's ChatGPT operator (agent), limited to browser interactions, struggled with reliability and failed to achieve significant user adoption—a cautionary reference point for Anthropic's broader ambitions.

What This Means

Anthropicis positioning Claude as an agent capable of independent action within user systems, not merely a conversational interface. This represents a meaningful shift toward autonomous AI assistants that operate directly in digital environments. However, the research preview designation and macOS-only rollout signal that Anthropic recognizes substantial work remains before this moves to production. The technology's success depends entirely on solving the reliability, security, and controllability problems that have challenged earlier automation features. The Vercept acquisition appears strategically timed to accelerate this development, though real-world performance data will ultimately determine whether desktop control becomes a standard feature or remains a specialized capability.