Claude gets a partner that works while you are away
Anthropic has confirmed the launch of Claude Duet, a proactive agent that monitors your work tools, decides what deserves attention and carries out tasks on its own while you are doing something else. Unlike the regular Claude, which waits for you to open a chat, Duet sends a morning brief, checks connected accounts every few hours and only reaches out when it judges something truly matters. It is the clearest move yet by Anthropic to turn a text assistant into an actual digital employee that acts in the background and reports back later.
Update: the US rollout is moving fast, outside the US stays unclear
Independent outlet TestingCatalog reported on October 28 that Anthropic is taking every US Claude Max subscriber off the Duet waitlist this week, with US Pro subscribers coming right behind. There is still no date for expansion beyond the US, and the outlet expects EU users specifically to wait considerably longer. Anthropic has not issued its own statement on this pace, so treat the timeline as a signal from a credible tracker rather than an official schedule.
How Duet decides what to do
The agent runs on Opus 5.5, Anthropic's most advanced model, and lives as a separate application for iOS, Android and the web, apart from the main Claude app during this preview. It inherits memory, skills and integrations already set up in the user's account and can trigger more than six thousand connectors from the Claude Marketplace. In practice that means scheduling meetings, building dashboards, drafting financial forecasts, preparing contracts and even handling week long projects inside one continuous thread, by text or voice.
Automatic purchases and the limits of trust
The trickiest part is money. Anthropic partnered with Stripe so Duet can make business purchases directly through Stripe Link, using virtual cards, and it also works with Shopify's Shop Pay. Anthropic says this is exactly the kind of action that requires explicit approval before it happens: the agent can read, draft and prepare anything on its own, but it stops before sending a message, making a payment, signing a document or deleting something. Trading securities and transferring money between accounts are off the table entirely. The credential architecture follows the design used for Claude Managed Agents, with logins and approvals running on a virtual machine isolated from the user's own device.
Why this matters now
Autonomous agents that act without constant supervision are the most contested race among the big AI labs right now, and each one is testing its own line between autonomy and control. Duet arrives as a direct answer to that appetite, but also as a test of how much money and sensitive data users are willing to hand over to an AI that works while they sleep. If the preview goes well, Anthropic gains a subscription product with far stronger retention than a plain chat, because the agent becomes part of the daily routine of whoever runs a small business.
What still needs to mature
Anthropic itself warns that Duet still makes mistakes: it can misread a thread, misjudge urgency and miss something that arrives between check ins. Personal errands are less developed than business ones, and for now there is no way for users to see the browser the agent uses to work behind the scenes. Access remains waitlist based, and background work counts against the same plan usage limits, which could hurt heavy daily Claude users.
Sources
Anthropic: official Claude Duet page; TestingCatalog: Anthropic opens Claude Duet preview for US Pro and Max users; TestingCatalog on X: Claude Duet rolling out to US Max users, October 28; Stripe: Claude Duet partnership announcement; Anthropic: Scaling Managed Agents.