Anthropic launches Claude Duet, a proactive AI agent, in research preview

Anthropic has announced Claude Duet, an AI agent aimed at professionals and small business owners, and is releasing it as a research preview. According to The Verge, it competes with Meta's Muse and OpenAI's Dots. It runs on Claude Opus 5.5 and is offered as a standalone app on iOS and Android, and on the web at claude.ai.

Anthropic describes a product that works in the background. Each morning it sends a short brief on the most important items across connected apps. During the day it monitors those apps and messages the user only when something needs attention. Users can also hand it tasks by text or voice, from scheduling meetings to building dashboards, financial or sales forecasts, prototypes and web apps.

Duet uses the user's existing Claude account, including memory, skills and connectors. It works on its own cloud computer, which has a virtual machine and a browser. Anthropic says it asks for approval before sensitive operations such as sending messages or paying, and that users can see what it has done and steer it.

Access is limited. Duet is on a waitlist for Claude Pro and Max subscribers in the US, and Team and Enterprise plans are not included yet. During the preview it shares a plan's usage limits with the rest of Claude, and its background checks count toward them. Larger jobs such as a financial model or a deck need more model turns and background work.

Most chat assistants wait for a prompt. Duet reverses that by deciding when to contact the user, which makes judgment about urgency the central product question. Anthropic acknowledges the risk, saying the agent may misread a thread or miscalibrate what is urgent, and that it can miss something that arrives between check-ins. It also says Duet is not as strong at personal errands such as booking tickets.

The safety design reflects the stakes of an agent that can handle payments and sensitive data. Anthropic says it added a dedicated classifier and a safety system prompt designed to reject requests to share the strictest tier of information, such as credit card numbers, and that approvals, logins and safety checks run in the agent's own cloud computer.

The launch also places Anthropic alongside Meta and OpenAI in a segment aimed at the workday of small businesses, where one person often handles hiring, finance, customers and operations. How well Duet performs in that role is not yet established, since the product is a preview with restricted access.

Watch whether Anthropic expands beyond US Pro and Max subscribers to Team and Enterprise plans, and when the waitlist opens up more broadly.

Also open is how the shared usage limits hold up, given that Anthropic says background checks and larger tasks consume more of a plan's allowance than a typical conversation.