Anthropic Launches Claude Duet Agent for Autonomous Background Tasks
Anthropic has launched Claude Duet, an AI agent built on Claude Opus 5.5 that operates in a dedicated VM and browser to execute cross-app tasks autonomously even when the user is offline. Featuring over 6,000 connectors and strict security approval protocols, the tool is currently in preview with waitlist access for US Pro and Max subscribers.
Full Report
In terms of capabilities, the application organizes important cross-app items daily, proactively reminds users of issues requiring attention, and continues to complete delegated tasks in the background after the user leaves. IT Home has attached a demo video below:
Claude Duet is built on Claude Opus 5.5 and is available as an iOS app, a standalone application, and via the claude.ai web version. A waitlist is initially open to Pro and Max subscribers in the United States; during the preview phase, it shares usage limits with other Claude features.
Unlike standard chatbots, Claude Duet possesses its own virtual machine and browser, allowing it to run continuously even when the user's device is off. Users can assign tasks via text or voice, such as scheduling meetings, organizing unread emails, tracking invoices, creating financial or sales forecasts, building dashboards, and developing web applications; upon task completion, it notifies the user or requests review.
Regarding the tool ecosystem, Claude Duet can access over 6,000 connectors and plugins from the Claude Marketplace, covering common office and business services such as Gmail, Slack, QuickBooks, Stripe, Gusto, LinkedIn, HubSpot, Canva, Shopify, and DocuSign. It can also complete online payments confirmed by the user via Link or Shop Pay, using virtual cards to prevent exposure of real card numbers.
In terms of security design, Anthropic states that Claude Duet does not directly access user passwords or login tokens; user approval is required before sending messages, making payments, signing documents, or deleting content. The system also includes specialized classifiers to intercept requests involving highly sensitive information such as credit card numbers and government ID documents.
However, Anthropic acknowledges that this remains an early version: Claude Duet may misinterpret conversations or judge the urgency of matters incorrectly; its ability to handle personal affairs is relatively weak, and it currently cannot show users its browser operation process.