BotBeat
...
← Back

> ▌

AnthropicAnthropic
PRODUCT LAUNCHAnthropic2026-03-23

Anthropic Launches Claude Computer Use: AI Agent Can Now Control Your Computer to Complete Tasks

Key Takeaways

  • ▸Claude can now autonomously control computers by using mouse, keyboard, and screen navigation to complete tasks when direct integrations aren't available
  • ▸The feature integrates with Dispatch to enable asynchronous task assignment from mobile devices, allowing users to delegate work and retrieve results across devices
  • ▸Anthropic emphasizes the research preview status and early maturity of the capability, with safeguards for security but acknowledged limitations for complex tasks
Sources:
Hacker Newshttps://claude.com/blog/dispatch-and-computer-use↗
Hacker Newshttps://twitter.com/i/status/2036195789601374705↗
X (Twitter)https://x.com/claudeai/status/2036195789601374705/video/1↗
Hacker Newshttps://xcancel.com/claudeai/status/2036195789601374705↗

Summary

Anthropic has introduced computer use capabilities to Claude Cowork and Claude Code, enabling Claude to autonomously control a user's computer by pointing, clicking, and navigating the screen to complete tasks. When Claude lacks access to direct integrations, it can open files, use browsers, and operate development tools without requiring setup. The feature is now available in research preview for Claude Pro and Max subscribers on macOS.

The computer use capability pairs seamlessly with Dispatch, Anthropic's recently launched mobile feature that allows users to assign Claude tasks from their phone and receive completed work on their desktop. This integration enables workflows such as automated morning briefings, code changes with PR submissions, and autonomous project management. Anthropic has implemented safeguards including prompt injection detection and permission-based access controls, though the company acknowledges the feature is early-stage and recommends starting with trusted applications and avoiding sensitive data.

  • Computer use is currently limited to macOS with Claude Pro and Max subscribers, marking a significant step toward AI agents that can operate within existing user workflows

Editorial Opinion

Claude's computer use capability represents a meaningful evolution in AI agent autonomy, transitioning from task execution via APIs to direct screen interaction. While the safety-first approach with permission gates and built-in safeguards is commendable, the acknowledged early-stage nature and macOS-only availability suggest Anthropic is rightfully cautious about releasing powerful automation tools. The combination with Dispatch creates a compelling use case for asynchronous work delegation, but success will depend on how reliably Claude handles real-world complexity across diverse applications.

Computer VisionGenerative AIMultimodal AIAI AgentsProduct Launch

More from Anthropic

AnthropicAnthropic
POLICY & REGULATION

Advanced AI Models Bring Government to 'Reflection Point,' CIA Official Says

2026-05-20
AnthropicAnthropic
RESEARCH

Anthropic Claude Code Sandbox Bypass: Second Vulnerability Exposes Critical Data Exfiltration Risk

2026-05-20
AnthropicAnthropic
RESEARCH

AI Safety Catastrophically Underfunded: Economic Model Reveals Incentive Gap

2026-05-20

Comments

Suggested

Google / AlphabetGoogle / Alphabet
PRODUCT LAUNCH

Google DeepMind Launches Gemini 3.5 Flash: New Lightweight AI Model

2026-05-20
Executive Office of the President of the United States (Policy/Regulation)Executive Office of the President of the United States (Policy/Regulation)
RESEARCH

SID Achieves Search Breakthrough with SID-1, Outperforming GPT-5 at 1k+ QPS Using Reinforcement Learning

2026-05-20
Helmholtz MunichHelmholtz Munich
RESEARCH

MouseMapper: AI Foundation Model Maps Systemic Damage from Obesity at Whole-Body Scale

2026-05-20
← Back to news
© 2026 BotBeat
AboutPrivacy PolicyTerms of ServiceContact Us