Claude Duet: Anthropic's proactive agent for small business

What we know

• What it is: according to Anthropic's announcement, a standalone app on iOS, Android and claude.ai, separate from the main Claude app, where the agent comes to you with what needs attention, completes tasks and keeps working when you step away. Anthropic says it built Duet around Claude Opus 5.5.
• Who and when: the research preview opens by waitlist, starting with US Pro and Max subscribers. It is included in those plans and uses the same usage limits.
• How it works: each morning it sends a short brief on the most important items across connected apps; during the day it checks them every few hours and only messages you when something needs you. It runs on a virtual machine with its own browser, so your laptop or phone doesn't need to be on.
• What it can reach: Anthropic cites more than 6,000 connectors and plugins in its marketplace. Examples in the announcement include replying to Gmail and Slack messages, finding unpaid invoices in QuickBooks or Stripe and drafting reminders, preparing payroll in Gusto, posting job listings to LinkedIn after approval, drafting follow-ups from HubSpot history, building Canva campaigns, and setting up, designing and managing a Shopify store.
• What stays with you: the announcement says that before it sends a message, makes a payment, signs a document or deletes anything, it shows exactly what it is about to do and waits for approval. Trading securities and moving money between accounts are off-limits even with approval.
• Purchases and sensitive data: it can buy things for the business with Link, Stripe's wallet, which pays with a virtual card after you approve the spend, and with Shopify's Shop Pay. Anthropic says a new classifier rejects requests to share card numbers and government IDs, and other sensitive data only goes out when you explicitly instruct it.
• Acknowledged limits: Anthropic warns it may misread a thread or miscalibrate urgency, that personal errands are weaker, and that it cannot yet show its browser. Background checks count toward plan limits, and big tasks use more.

What changes and what doesn't

For now, nothing changes for anyone outside the US or without a Pro or Max plan: the preview starts there, by waitlist. Anthropic has not given dates for other countries.

What does change is the working model: instead of opening a chat and asking for something, the agent reviews your inbox, billing or store on its own and proposes actions. The key question for a business shifts from what to ask the AI to what to let it into and what it may do without asking.

How to tell if it fits and what to prepare today

1. Map where your business lives. Email, billing, online store, CRM, social accounts. An agent like this is only useful where a connector exists; if your site or store is not on a connectable platform, the agent can't reach it.
2. Decide what you would not delegate. Before connecting anything, separate what the agent may read, what it may draft and what only you do. The announcement says it asks before sending, paying, signing and deleting; editing a store or posting to social media also deserves a review before it goes live.
3. Be careful with changes to your site. If you connect it to a store such as Shopify, every change to product pages, prices or design affects what customers and Google see. Check copy and links before approving.
4. Budget the usage. Because it shares plan limits and checks accounts every few hours, tell it what to watch and how often so background work doesn't drain your plan.
5. If you're outside the US, there is nothing to switch on yet. You can still tidy up access: business accounts separate from personal ones and least-privilege permissions in each tool, which helps with this agent or any other.

Related: Gemini Agent: Google Gives AI Its Own Workspace Account

Sources

• Anthropic, "Introducing Claude Duet"

Updates: if Anthropic opens Claude Duet to more countries or plans, it will be added here with a link.