Claude Duet

Claude Duet is a research preview from Anthropic: a proactive agent you message by text or voice. It works in the apps you connect, keeps going on its own cloud computer after you close the app, and comes back with decisions or finished work. Anthropic says it built Claude Duet for professionals and small business owners with more on their plates than they can track.

What It Is

Claude Duet is a standalone messaging app that runs on your Claude account, so your memory, skills, and connectors are available from the first message. The regular Claude app waits for you to open a chat. Claude Duet comes to you: it sends a morning brief based on your email and calendar, drafts replies for you to review, and follows up on threads that have gone quiet. Anthropic built it around Claude Opus 5.5 and describes its main strength as judgment about what needs you, what can wait, and when to check in first.

How the Workflow Runs

You keep one main thread and can open separate channels for different topics that share the same memory. You describe a task in a short message, and it decides which apps to use and how much effort the job needs, then works in the background on a virtual machine with its own browser, so your phone or laptop does not need to stay on. It checks your connected accounts every few hours, and you can tell it what you want to hear about and how often.

A Tasks tab shows what it is working on and which tasks need your input; it may suggest a task is finished, but only you can mark it complete. Larger jobs come back as files you can review and edit, such as a research brief, financial model, slide deck, or dashboard, and they appear in a Make tab. It can search your past Claude Duet chats and Claude chats outside projects for context, and you can report problems by describing them in the chat.

Business Tasks and Integrations

Claude Duet can use the connectors and plugins in the Claude Marketplace, which Anthropic says number more than 6,000, and it can suggest plugins such as a small business plugin. Documented examples include finding missed Gmail and Slack messages, checking QuickBooks or Stripe for unpaid invoices and drafting reminders, preparing payroll in Gusto, preparing agreements for signature in Docusign, drafting follow-ups from HubSpot deal history, turning a launch into a Canva campaign, posting an approved job listing to LinkedIn, and setting up a Shopify store. For sites with no connector, it uses its cloud browser. It can buy things for you through Link, Stripe's digital wallet, or check out with Shop Pay, and each approved purchase uses a one-time card.

Anthropic shares accounts from early testers: a real estate broker whose calendar error it flagged without being asked, a restaurant owner with about 110 employees who used it for invoices, accounting, and customer support, and a brand strategist whose call it offered to move because it clashed with a flight.

Approvals and Safety

Claude Duet reads, drafts, and prepares work on its own, but it is designed to show you exactly what it will do and wait for approval before it sends a message, makes a payment, signs a document, or deletes anything. If you cannot approve right away, it finishes the rest of the task and brings the approval back to you later. It cites which email, file, or spreadsheet a number came from.

Anthropic says the product uses its auto mode classifier to check actions, plus a dedicated classifier and safety prompt that refuses to share credit card numbers or government ID and shares health or financial details only when you explicitly tell it to. Passwords are typed into a separate form that goes straight to the cloud browser, so they stay out of the chat. Anthropic also says red teams and an outside firm penetration-tested the product. The documentation notes that prompt injection in emails or web pages is a real risk and that no defense is perfect, which is why consequential actions need your approval.

Current Status and Limits

Claude Duet is a research preview, access opens in waves through a waitlist, it is US-only, and features can change. You cannot choose a model, watch or take over its browser, or have it see local files or change your work in Claude. Because it checks accounts at intervals, urgent items can arrive between checks, and it can misread a thread or misjudge urgency. It does not make phone calls or text other people, some actions such as trading securities, moving money between accounts, or accessing someone else's account are off-limits even with approval, Slack works with limits, and Microsoft 365 and Personal Finance are not yet available. Anthropic also says personal errands such as booking tickets are weaker than work tasks for now. It is available as an iOS and Android app and on the web at claude.ai.