Illustration of Claude AI controlling a Mac desktop, with open apps like Slack and Calendar, highlighting new research preview features.
Illustration of Claude AI controlling a Mac desktop, with open apps like Slack and Calendar, highlighting new research preview features.
AI 生成的图像

Anthropic's Claude AI Gains Full MacOS Desktop Control in Research Preview

Building on its January Cowork feature, Anthropic has launched a research preview for Claude Code and Cowork tools, enabling Pro and Max subscribers' Claude AI to directly control Mac desktops—pointing, clicking, scrolling, and navigating screens for tasks like opening files, using browsers, developer tools, and app interactions such as Google Calendar and Slack. Safeguards address security risks, amid competition from tools like OpenClaw.

Anthropic announced on March 24, 2026, an expansion of its Claude AI capabilities, following the January launch of Cowork—which initially enabled folder access for office tasks. Now, Claude Pro and Max subscribers on MacOS can grant the AI full screen control in a 'research preview.' When direct app Connectors (e.g., for Google Calendar or Slack) are unavailable, Claude simulates keyboard and mouse inputs to scroll, click, retrieve/send files, navigate browsers, run developer tools, and more—always requiring explicit user permission, which is revocable at any time.

The feature integrates with Anthropic's Dispatch tool for remote or phone-based task assignment, such as morning email checks or briefings, provided the target computer remains powered on. While functional, Anthropic notes it is slower and more error-prone than Connectors, may require retries for complex tasks, and 'won't always work perfectly.'

Robust safeguards include protections against prompt injection, vulnerability scans, and default blocks on sensitive apps like investment platforms or cryptocurrency tools. Claude is trained to avoid risky actions such as moving money, modifying files, scraping facial images, or entering sensitive data. However, these are 'not perfect' or 'absolute,' and the AI can view on-screen content—including personal data—prompting warnings to use trusted apps only and avoid sensitive information. Experts highlight agentic AI risks like unintended actions or hijacking.

This follows similar tools including open-source OpenClaw (whose creator Peter SteinBerger was hired by OpenAI), Nvidia's NemoClaw, Perplexity’s Personal Computer, and Manus’s My Computer. Anthropic seeks user feedback to refine the preview.

人们在说什么

X discussions on Anthropic's Claude AI gaining full MacOS desktop control via Cowork and Claude Code show excitement over enhanced productivity and agentic features, but skepticism regarding speed, Mac-only availability, and security vulnerabilities like prompt injection and expanded attack surfaces.

相关文章

Professional using Anthropic's Claude AI Cowork feature on MacBook to automatically organize files and generate reports, as shown in a realistic office scene.
AI 生成的图像

Anthropic launches Cowork feature for Claude AI

由 AI 报道 AI 生成的图像

Anthropic has introduced Cowork, a new tool that extends its Claude AI to handle general office tasks by accessing user folders on Mac computers. Designed for non-developers, it allows plain-language instructions to organize files, create reports, and more. The feature is available as a research preview for Claude Max subscribers.

On February 5, 2026, Anthropic and OpenAI simultaneously launched products shifting users from chatting with AI to managing teams of AI agents. Anthropic introduced Claude Opus 4.6 with agent teams for developers, while OpenAI unveiled Frontier and GPT-5.3-Codex for enterprise workflows. These releases coincide with a $285 billion drop in software stocks amid fears of AI disrupting traditional SaaS vendors.

由 AI 报道

Anthropic has introduced a beta feature allowing its Claude AI to generate interactive charts, diagrams, and visualizations directly in chats. The tool aims to enhance understanding by replacing dense text with visuals when appropriate. It is available to all users on web and desktop platforms.

OpenAI has released a dedicated macOS application for its Codex AI coding tool, enhancing its capabilities to manage multiple AI agents for complex tasks. The app builds on Codex, which debuted last spring as a response to competitors like Anthropic's Claude Code. It introduces features like Skills and Automations to streamline workflows for developers.

由 AI 报道

Apple has released Xcode 26.3, enhancing its integrated development environment with deeper integration for AI coding agents like Claude and OpenAI's Codex. The update leverages the Model Context Protocol to allow these agents to access documentation, file structures, and project settings more effectively. This builds on AI features introduced in Xcode 26 at WWDC 2025.

AI coding agents from companies like OpenAI, Anthropic, and Google enable extended work on software projects, including writing apps and fixing bugs under human oversight. These tools rely on large language models but face challenges like limited context processing and high computational costs. Understanding their mechanics helps developers decide when to deploy them effectively.

由 AI 报道

Anthropic's recent update to its CoWork platform has led to significant market reactions in the software industry. The U.S. software sector saw a widespread sell-off, losing over $1 trillion in value, according to Fortune. This development highlights investor uncertainty around AI-native workflows and their impact on SaaS stocks.

 

 

 

此网站使用 cookie

我们使用 cookie 进行分析以改进我们的网站。阅读我们的 隐私政策 以获取更多信息。
拒绝