Claude Duet: Anthropic’s Proactive AI Agent App Explained

Claude Duet is a new app from Anthropic in which Claude works as a proactive AI agent. Instead of waiting for prompts, it surfaces what needs your attention, completes tasks in the background, and asks for approval before it acts for you. A research preview opened on October 21, 2026 for Pro and Max subscribers in the US, via a waitlist.

Anthropic is aiming the app at professionals and small business owners who are juggling more than they can comfortably track. Below is what Claude Duet does, how it works under the hood, the safety controls it ships with, and what to know before you sign up.

What is Claude Duet?

Claude Duet is a standalone app for iOS and Android, also available on claude.ai. During the research preview it is included with Claude Pro and Max plans and draws from the same usage limits as the rest of Claude. There is no separate price.

Its core idea is a shift in who starts the conversation. A typical chatbot waits for you. Claude Duet comes to you first: each morning it sends a short brief on the most important items across your connected apps, and during the day it checks in only when something actually needs you. The app runs on Claude Opus 5.5, and Anthropic says the product's main selling point is judgment, deciding what needs you, what can wait and when to check before acting.

Key features of Claude Duet

A morning brief and proactive alerts

Claude Duet scans your connected accounts every few hours rather than constantly. Anthropic chose that interval because nonstop checking would burn through plan limits much faster. You can tell it what you want to hear about and how often, and the company says it is still tuning when the agent decides to message you.

Tasks that keep running when you step away

Each Claude Duet instance has its own virtual machine with a browser, so your laptop or phone doesn't need to stay on. You describe a task in a short message, it works out which apps and steps are needed, and it notifies you when the job is done or needs review. Tasks range from scheduling meetings to building dashboards, financial or sales forecasts, prototypes and web apps.

One thread, text or voice

All interaction happens in a single ongoing thread, by typing or by voice. A quick reply and a week-long project start in the same place. The app also inherits your memory, skills and connected apps from your existing Claude account, so setup is light.

Connectors for business tools

Claude Duet can use the more than 6,000 connectors and plugins in the Claude Marketplace. Examples Anthropic gives include:

• Finding missed Gmail and Slack messages and helping you reply
• Checking QuickBooks or Stripe for unpaid invoices and drafting reminders
• Preparing and running payroll in Gusto
• Drafting job posts and publishing to LinkedIn once you approve
• Preparing agreements and sending them for signature through Docusign
• Writing follow-up emails from HubSpot deal history
• Turning a product launch into a Canva social campaign
• Setting up and managing a Shopify store

It can also make business purchases through Link, Stripe's digital wallet, which pays with a virtual card so your real card number is never shared, and it can check out with Shop Pay. Sign-in through saved 1Password logins is listed as coming soon.

How does Claude Duet keep you in control?

An agent that reads your email and acts across your accounts raises obvious trust questions. Anthropic's answer has two layers: approval gates for consequential actions, and a stack of security systems borrowed from its other agent products.

Approval before high-stakes actions

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 you exactly what it intends to do and waits for a yes. Some actions are blocked even with your approval, such as trading securities or moving money between accounts. You can disconnect any app at any time.

Security systems behind the agent

• Trustworthy agents framework: the same framework that guides Claude Code and Claude in Chrome.
• Credential handling: modelled on Claude Managed Agents, so the agent never directly sees your passwords or sign-in tokens.
• Auto mode classifier: approves actions and guards against steps you didn't ask for.
• Prompt-injection defenses: drawn from Anthropic's browser-use research.
• Adversarial testing: internal red teams and an external firm penetration-tested the app, and it falls under Anthropic's public bug bounty program.

A classifier for sensitive information

Claude Duet also launches with a dedicated classifier and safety system prompt that inspects each outgoing message. It refuses to share the most sensitive tier of data, including credit card numbers and government IDs. Other sensitive material, such as health or personal finance details, goes out only when you explicitly tell the agent to send it.

How early testers use Claude Duet

Inside Anthropic, Claude Code lead Boris Cherny talks through his to-do list with the agent while walking instead of saving it for his desk. Chelsea Larsson, who leads model and product behavior design, hands over her list before spending time with her kids and gets an evening report. The Duet team, led by Karan Ahuja, uses the app to schedule user interviews, prototype and ship code.

Small business testers report similar patterns. A real estate broker had the agent flag a property showing booked on the wrong day without being asked. A restaurant owner with around 110 employees uses it for accounting, customer support, marketing and store-by-store invoice reviews. A brand strategist found it had prepared a client call brief and offered to move the call because it clashed with her flight.

Limitations of the research preview

• It can misjudge urgency or misread a thread; this is an early build.
• Work comes first. Personal errands like booking tickets are weaker.
• No browser view yet. You can't watch its browser as it works.
• Separate app. It lives apart from the main Claude app for now.
• Usage adds up. Background checks count toward your plan limits, and big jobs like a financial model or a deck use more.
• US only, Pro and Max only at launch.

How to get access to Claude Duet

Join the waitlist on the Claude Duet product page. Anthropic will email you when your access is ready. You'll need an active Claude Pro or Max subscription in the US.

Source: Introducing Claude Duet, Anthropic, October 21, 2026.