Anthropic、通常のアプリ連携が不十分な場合にClaudeがデスクトップを直接操作できる新機能を発表

Update, April 3, 2026:

The feature is now available to Pro and Max users for Claude Code and Claude Cowork in the Windows app.

Original article from March 24, 2026:

Anthropic has announced a new feature for its AI assistant Claude: the ability to directly operate a user's computer, handling tasks people would normally do themselves at their desk.

According to Anthropic, Claude first tries existing integrations like Slack, calendars, and other connected apps. It only takes direct control of the desktop when no other interface is available, making it a fallback, not the default. The feature ships as a research preview in Claude Cowork and Claude Code and is currently limited to macOS.

It's a step for Anthropic toward building an AI agent that actively works inside a user's digital environment instead of just generating text. Anthropic also rolled out a feature called "Dispatch" alongside Cowork's computer control, letting users remotely control their own computer from anywhere.

Vercept AI acquisition laid the groundwork for desktop control

The feature is partly built on technology from Anthropic's acquisition of Vercept AI, a startup focused on AI-powered computer control. According to co-founder Kiana Ehsani, her team shipped its first product less than four weeks after joining Anthropic.

Before the acquisition, everyone involved worried it would slow things down. The opposite happened, Ehsani says. She credits Anthropic's culture: "Everyone moves fast, everyone is incredibly smart, humble and supportive, and it's really easy to get things done," she writes on X. In her view, Anthropic's biggest competitive advantage is its people.

Full computer use by AI still has a lot to prove

Whether computer use can prove reliable and secure enough for real-world work beyond its research preview status is still an open question. Giving an AI control over an entire desktop raises serious concerns around data privacy, error rates, and controllability, things Anthropic would likely need to address before a broader rollout.

Similar features limited to just the browser haven't been reliable enough to gain real traction, as OpenAI found with its ChatGPT operator, aka agent. Anthropic is taking a much more ambitious swing with full desktop control, but that also means a significantly larger attack surface.

AI News Without the Hype – Curated by Humans

Subscribe to THE DECODER for ad-free reading, a weekly AI newsletter, our exclusive "AI Radar" frontier report six times a year, full archive access, and access to our comment section.

Subscribe now