On October 21, 2026, Anthropic opened a research preview of Claude Duet, according to Unite.AI. The company describes the app as an agent that surfaces items needing attention, completes requested tasks, keeps working when you step away, and checks before acting on your behalf. Access runs through a waitlist and starts with Pro and Max users in the US. Anthropic said it will email waitlisted users once their access is ready.
How it works
Anthropic said it built Claude Duet around Claude Opus 5.5. Each morning the app sends a short brief on the most important items across your connected apps; during the day it monitors those apps and messages you only when something needs attention. Tasks range from scheduling and dashboards to financial or sales forecasts, prototypes and web apps. It carries over memory, skills and connected apps from your Claude account and runs on a virtual machine with its own browser, so your laptop or phone does not need to be on. It notifies you when a task is complete or needs review.
Approvals and payments
According to Anthropic, 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 exactly what it intends to do and waits for approval. Some actions are off-limits even with approval, such as trading securities and moving money between accounts. The app can use more than 6,000 connectors and plugins in the Claude Marketplace. For purchases it works with Link, Stripe's digital wallet, and pays through a virtual card so the real card number is not shared.
Safety and early testers
Anthropic says the safety design follows its framework for trustworthy agents, and the app never directly sees passwords or sign-in tokens. A dedicated classifier is designed to reject requests to share the most sensitive tier of information, such as credit card or government ID numbers; other sensitive data is sent only on explicit user instruction. According to Anthropic, red teams and an external firm penetration-tested the app. Early testers include Anthropic employees and some small business owners and entrepreneurs. Anthropic itself describes the preview as an early version that can still misread a thread or miscalibrate urgency, and says personal errands such as booking tickets are weaker because the core functionality was built around work.
Assessment
For companies in Germany, the approval gates before sending, paying or signing may be relevant. The sources leave open how data protection and access rights for connectors are handled in practice, and whether availability outside the US is planned. Anthropic has said it will publish more about its safety design in the coming days.
Sources
Editorially owned by Ideal Syka. Sources and method: Newsroom & method. Tips and corrections: ai@i6eal.de.
