Anthropic's Claude Duet is an agent that acts for you, waitlist only

Anthropic launched Claude Duet on October 21 as a research preview. It is a standalone app on iOS and Android, plus claude.ai, built on Claude Opus 5.5. Duet sends a short morning brief across your connected apps, messages you only when something needs attention, and takes tasks from scheduling meetings to building dashboards, forecasts and web apps. It carries over your Claude memory, skills and connected apps, works in its own virtual machine with a browser, and plugs into 6,000+ connectors like Gmail, Slack, QuickBooks, Stripe, HubSpot, Shopify and Docusign. Before it sends a message, makes a payment, signs a document or deletes anything, it shows you the action and waits for your approval. Purchases run through Link or Shop Pay with your explicit approval, and some actions, like trading securities or moving money between accounts, are off-limits entirely.

This is Anthropic's answer to Meta's Muse and OpenAI's Dots, and the clearest sign yet that the big AI labs are moving from chatbots that answer to agents that act. The audience for it is exactly the professionals and small business owners drowning in admin. The approval-before-action model is the part to watch: it keeps a human in the loop while still letting the agent work unattended.

You cannot just try it. Access is a waitlist, starting with Pro and Max subscribers in the US only, so there is no free way in, and Duet shares your plan's usage limits, with its background checks burning through them. Anthropic itself calls it an early version that can misread a thread or misjudge what is urgent, says it is weaker at personal errands like booking tickets, and it cannot yet show you its browser. It also lives in its own app, separate from the main Claude app.