Anthropic on October 21, 2026 opened a research preview of Claude Duet, an app the company said turns Claude into a proactive agent that surfaces items needing attention, completes requested tasks, keeps working when users step away, and checks before acting on their behalf. Access runs through a waitlist, starting with Pro and Max users in the US.
Anthropic said it built Claude Duet for professionals and small business owners with more on their plates than they can track. The app is available as a standalone app on iOS and Android and on claude.ai; during the research preview it is included in Pro and Max plans and uses the same usage limits as the rest of Claude. Anthropic said it will email waitlisted users as soon as their access is ready.
How Claude Duet Works
Anthropic said it designed Claude Duet around Claude Opus 5.5 and built it to decide what needs the user, what can wait, and when to check first. Each morning the app sends a short brief on the most important items across the user’s connected apps; during the day it keeps tabs on those apps and messages only when something needs attention. Tasks range from scheduling meetings to building dashboards with dozens of tools, creating financial or sales forecasts, prototyping, and building web apps.
Claude Duet carries over the memory, skills, and connected apps from a user’s Claude account, and users message it by typing or voice in one ongoing thread. The app runs on a virtual machine with its own browser, so it does not need the user’s laptop or phone to be on; it sends a notification when a task is complete or needs review. Anthropic said the app checks connected accounts every few hours because nonstop checking would consume plan limits much faster; those background checks count toward the shared usage limits, and larger tasks such as building a financial model or a deck require more model turns.
Approval Gates, Connectors, and Payments
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 exactly what it is about to do and waits for approval. A few actions are off-limits even with approval, including trading securities and moving money between accounts, and users can disconnect any app Claude Duet can access.
The app can use the more than 6,000 connectors and plugins in the Claude Marketplace, with Anthropic’s listed uses spanning help finding and replying to missed Gmail and Slack messages, unpaid-invoice checks in Intuit QuickBooks or Stripe with drafted reminders, preparing and running payroll in Gusto with hours, bonuses, and reimbursements filled in, and LinkedIn job postings published after the user edits and approves them. Other listed uses include routing agreements through Docusign for review and signature, drafting follow-up emails from HubSpot contact and deal history, turning a product launch into a Canva social campaign, and setting up, designing, and managing a Shopify store, with a richer visual storefront editor coming soon.
For business purchases, Claude Duet works with Link, Stripe’s digital wallet: the app confirms the item with the user, Link asks the user to approve the spend, and payment runs through a virtual card so the real card number is not shared. It can also use Shop Pay from Shopify to check out at online stores, and Anthropic said that, with user permission, it will soon be able to sign in to sites with logins saved in 1Password.
Safety Design and Testing
Anthropic said Claude Duet draws on its framework for trustworthy agents, which also guides Claude Code and Claude in Chrome, and that its credential handling follows the design used for Claude Managed Agents, so the app never directly sees user passwords or sign-in tokens. The company said actions are approved by the auto mode classifier, the same safeguard that protects claude.ai and Claude Code from taking steps users did not request, and that the app’s defenses against hidden instructions and prompt injection come from its browser-use research. Anthropic said red teams and an external firm penetration tested Claude Duet, and the app is covered by its public bug bounty program on HackerOne.
Anthropic released Claude Duet with a new dedicated classifier and safety system prompt that checks each message, which the company said is designed to reject requests to share the strictest tier of information, including credit card numbers and government identification. Other sensitive information, such as health or personal financial details, passes only when the user has explicitly instructed Claude Duet to send it, Anthropic said. The company said it will share more details about the app’s safety design in the coming days.
Early Testers and Stated Limitations
Anthropic said its employees have been using the app. Boris Cherny, who leads the Claude Code team, talks through his mental to-do list with Claude Duet on walks; Chelsea Larsson, who leads the model and product behavior design team, hands it her to-do list before leaving to be with her children, and it saves the results for an evening report; and Karan Ahuja, who leads the Claude Duet team, uses it to schedule user interviews, partner on prototypes, and ship code.
Anthropic also gave small business owners and entrepreneurs early access. Jess Provencher, broker and owner of Pro Homes, said the app flagged, without being asked, a property showing booked on her calendar for the wrong day. Amin Ebrahimi, owner of Shahs of Kabob in South Florida, which has about 110 employees, said he has used it for accounting, customer support, and marketing, working through invoices store by store. Mya Dalton, a brand strategist, said Claude Duet prepared a brief for an upcoming call with a potential client and, seeing the call fell just before her flight, offered to move it.
Anthropic described the preview as an early version that can still misread a thread or miscalibrate what is urgent, and said personal errands such as booking tickets are not as strong because the app’s core functionality was built around work. Claude Duet cannot yet show users its browser and, for now, lives in its own app separate from the main Claude app; Anthropic said what it learns from early users will shape how the two work together, and that it will keep tuning when the app messages users based on tester feedback.
Credit: Source link
