The AI race has taken an interesting turn lately, with frontier labs now competing to build agents that can keep working even when you are not actively talking to them. Meta helped push that idea into the spotlight with Muse, OpenAI followed soon after with Dots, and now Anthropic is throwing its hat in the ring with Claude Duet.

Duet is launching as a research preview for Pro and Max subscribers in the US through a waitlist. Anthropic is pitching it mainly at professionals and small-business owners who want help keeping track of work without constantly checking every app themselves.

Duet wants to handle the boring stuff first

Each morning, Claude Duet can put together a brief covering the things it thinks deserve your attention. It can then keep monitoring your services throughout the day and message you when something needs a decision.

Duet is also capable of performing more complex tasks than serving simple reminders or summaries. It can also schedule meetings, build dashboards, prepare financial or sales forecasts, draft follow-up emails, and even work on web apps. And since it runs on its own virtual machine and browser, those tasks can continue even when your computer is off.

Claude Duet still needs you looking over its shoulder

Anthropic is giving Duet quite a bit of freedom, but clear limits remain on what it can do on its own. Things like sending messages, making payments, signing documents, or deleting anything still require the user’s permission.

Duet is still a research preview, and Anthropic has acknowledged that the agent still has a lot to learn. The company says it can misread a thread, misjudge the urgency of something, or miss an update that arrives between check-ins, and all that background work also counts toward the same usage limits as the rest of Claude. Bigger jobs, such as building a financial model or a presentation, can burn through those limits faster because they require more model turns.