Anthropic Claude Will Now Take Over Your Computer While You’re on Your Phone

TLDR

• Anthropic launched a feature letting Claude control a Mac computer to complete tasks remotely
• Users send a task from their phone; Claude opens apps, browses the web, and fills spreadsheets
• The feature works alongside Dispatch, which lets users assign tasks to Claude via phone
• Anthropic warns the tool is still early-stage and suggests avoiding apps with sensitive data
• The move comes as OpenClaw, a rival agentic AI framework, went viral and triggered a race among tech firms

Anthropic has launched a new feature that lets its Claude AI take control of a user’s computer to complete tasks. The update was announced on Monday and is available now for Claude Pro and Claude Max subscribers.

The feature works with MacOS only. Users send a prompt from their phone, and Claude carries out the task on their desktop — opening apps, navigating a browser, filling in spreadsheets, or moving files.

Anthropic showed a demo where a user running late for a meeting asked Claude to export a pitch deck as a PDF and attach it to a calendar invite. Claude completed the task without further input.

The tool connects to apps like Google Calendar and Slack through built-in connectors. If a connector isn’t available, Claude can use the keyboard and mouse manually to get the task done.

Claude will always ask for permission before accessing a new app. Users can also stop any task mid-way through.

How It Works With Dispatch

The computer use feature is built to work with Dispatch, a tool Anthropic released last week inside Claude Cowork. Dispatch lets users have a continuous back-and-forth with Claude from any device and assign it ongoing tasks.

Together, the two features can handle things like sending a morning briefing, running code tests, or checking emails automatically.

Anthropic says complex tasks may not always work on the first try. The company is calling this a research preview to gather feedback and identify where improvements are needed most.

Security Risks and Warnings

Anthropic has put safeguards in place to reduce risk, including automatic scanning for prompt injection attacks. But the company still warns users to avoid using the feature with apps that handle sensitive data — some of those apps are disabled by default.

Security experts have flagged broader concerns about agentic AI. These tools can take large actions quickly and with little warning. There is also a risk of bad actors hijacking the agent to access personal files or systems.

Anthropic said it will keep updating its safeguards as new threats emerge.

The update puts Anthropic in more direct competition with OpenClaw, an open-source framework that went viral earlier this year. OpenClaw allows users to send tasks via WhatsApp or Telegram and run them locally on a device. Nvidia last week launched NemoClaw, an enterprise version of the framework.

OpenAI also moved into this space last month, hiring Peter Steinberger, the creator of OpenClaw, to work on personal AI agents.

As of now, Claude’s computer use feature remains in research preview on MacOS for Pro and Max subscribers.

The post Anthropic Claude Will Now Take Over Your Computer While You’re on Your Phone appeared first on CoinCentral.