An AI assistant that reports on its own each morning which emails are sitting unanswered and which invoice is overdue, without anyone having to ask. That is exactly what Anthropic promises with the new research preview of Claude Duet, released on October 21, 2026. Instead of waiting for input, the app independently checks connected accounts and reaches out only when something genuinely needs attention.
Table of Contents
What Claude Duet Actually Is
Claude Duet is a new standalone app in which Claude works as a proactive agent rather than a purely reactive chatbot. It is built around the Claude Opus 5.5 model and is designed specifically for professionals and small business owners with more on their plates than they can track. The app is available on iOS and Android, as well as through claude.ai.
Access during the research preview is limited to a waitlist, available only to US-based subscribers on the Pro or Max plans. Team and Enterprise plans are explicitly not included in this initial rollout.
How the Daily Briefing and Monitoring Work
Each morning, Claude Duet sends a short briefing covering the most important items across a user’s connected apps, such as email and calendar. Throughout the day, it checks those accounts every few hours and only sends a message when something actually requires a decision. According to Anthropic, continuous real-time monitoring is deliberately avoided, since it would burn through a plan’s usage limits far too quickly. Users can also configure what they want to be notified about and how often.
A Cloud Computer That Keeps Working
Claude Duet runs on its own virtual machine with its own browser in the cloud, so tasks keep running even after a laptop or phone is switched off. Communication happens in one ongoing thread by text or voice, and the app automatically carries over memory, skills, and already-connected apps from a user’s existing Claude account, keeping setup minimal.
From Simple Chores to Complex Projects
Users can hand Claude Duet both simple administrative tasks, such as scheduling meetings, and more complex jobs, including building dashboards that pull from multiple tools, creating financial or sales forecasts, prototyping, and building web apps. Finished results appear in the app’s “Make” tab for review.
Over 6,000 Connectors From the Claude Marketplace
Claude Duet can tap into more than 6,000 connectors and plugins available in the Claude Marketplace, including Gmail, Slack, LinkedIn, Canva, HubSpot, QuickBooks, Stripe, Gusto, Docusign, and Shopify.
Concrete examples cited by Anthropic include: finding unanswered Gmail and Slack messages, tracking down unpaid invoices in QuickBooks or Stripe and drafting reminders, preparing payroll runs in Gusto, drafting job postings and publishing them to LinkedIn after approval, preparing contracts for signature through Docusign, drafting follow-up emails based on HubSpot deal data, building social media campaigns in Canva, and setting up and managing a Shopify store.
Where the Agent Draws the Line
Claude Duet reads, drafts, and prepares work on its own, but before it sends a message, makes a payment, signs a document, or deletes anything, it shows the user the exact planned action and waits for explicit approval. Certain actions remain off-limits even with approval, including trading securities and moving money between accounts. Users can disconnect any connected app at any time.
Security and Privacy Safeguards
Claude Duet never directly sees passwords or login tokens for connected accounts; credential handling follows the same design used for Claude’s other managed agents. Its actions are approved by the same “Auto Mode Classifier” that protects Claude Code, and its defenses against hidden instructions and prompt injection draw on Anthropic’s browser-use research.
A new dedicated classifier and safety system prompt screen every outgoing message and are designed to fully block the sharing of highly sensitive data such as credit card numbers or government identification numbers. Health or personal finance details are only shared when the user has explicitly instructed Claude Duet to do so. Red teams and an external security firm penetration-tested Claude Duet, and the app falls under Anthropic’s public bug bounty program.
Still an Early Version
Anthropic explicitly describes Claude Duet as an early release that may misread a thread or miscalibrate what is urgent. Personal errands, such as booking tickets, do not work as well yet as work-related tasks, and users cannot currently watch the agent’s browser in real time.
Anthropic’s Answer to Meta’s Muse and OpenAI’s Dots
Press coverage from outlets including CNBC, VentureBeat, and MacRumors frames Claude Duet explicitly as Anthropic’s answer to similar proactive agent products already on the market: Meta’s Muse and OpenAI’s Dots. During the research preview, Claude Duet shares the same usage limits as the rest of a user’s Claude account – its background checks and larger tasks count against those existing limits, and there is no separate pricing or dedicated API for Duet yet.
Anyone wondering what fundamentally sets AI agents apart from classic chatbots can find an accessible explanation in our foundational guide What Are AI Agents? AI Agents vs Chatbots Explained. For a broader look at competing tools, see our AI Agent Tools Compared.
Source: Anthropic, CNBC, VentureBeat, MacRumors.
