Anthropic Brings Computer Use to macOS as Claude Gains Ability to Control Desktop Apps Autonomously
Anthropic has launched a research preview enabling Claude to operate Mac desktops directly, opening files, navigating browsers, and executing tasks across applications with a permission-based safety model, available to Pro and Max subscribers.
Overview
Anthropic has launched computer use capabilities for Claude on macOS, enabling the AI assistant to take direct control of a Mac’s mouse and keyboard to complete tasks autonomously. The feature, released as a research preview on March 23, 2026, is available to Claude Pro and Claude Max subscribers through both Claude Code and Claude Cowork, according to 9to5Mac.
The release marks a significant step in the race to build AI agents capable of performing real work on personal devices, moving computer use from API-level developer tooling into consumer-facing desktop applications.
How It Works
When enabled on macOS, Claude can point, click, and navigate like a human user, opening and editing files, browsing the web, running development tools, and handling multi-step software tasks, as reported by Engadget. The system prioritizes direct connectors to supported services such as Google Workspace and Slack when available, falling back to screen-level interaction when dedicated integrations do not exist.
Claude Code, Anthropic’s tool aimed at programmers, and Claude Cowork, a broader productivity tool introduced in January 2025 for non-technical users, both gain the computer use capability with this update, according to Engadget.
Dispatch: Remote Task Assignment
The update also integrates with Dispatch, a feature released the prior week that allows users to assign Claude tasks from an iPhone while the AI operates on the Mac in the background. Anthropic demonstrated a use case in which Claude exported a pitch deck as a PDF and attached it to a meeting invite while the user was away from their desktop, according to 9to5Mac.
Safety and Permissions
Anthropic has adopted a permission-first approach to the feature. Claude requests user approval before accessing each new application and users retain the ability to halt execution at any point, as reported by 9to5Mac. The company acknowledged limitations in a statement reported by the same outlet: “Claude can make mistakes, and while we continue to improve our safeguards, threats are constantly evolving.” Anthropic recommended that users “start with the apps you trust and not working with sensitive data.”
Anthropic also advises against using the feature to handle sensitive information during the research preview phase, signaling that safety protocols remain under active development, according to Engadget.
Pricing and Availability
The computer use feature is available on macOS to Claude Pro subscribers starting at $17 per month and Claude Max subscribers at $100 or $200 per month. The release is limited to macOS for now, in contrast to competitors such as OpenClaw, which supports macOS, Windows, and Linux, as noted by 9to5Mac.
Competitive Landscape
The launch intensifies competition in the AI agent market. OpenAI introduced native computer use capabilities in GPT-5.4 earlier in March 2026. Meta’s Manus platform includes a My Computer feature for Mac, and Perplexity offers a cloud-based AI agent running on Mac mini hardware, as reported by 9to5Mac. The underlying technology draws from the same lineage as OpenClaw, the open-source AI agent framework that went viral on GitHub earlier in 2026, according to 9to5Mac.
Anthropic’s approach differs from OpenClaw’s plugin-based architecture by offering tighter security controls through its permission system, positioning safety as a differentiator in a rapidly expanding market for autonomous desktop agents.