Anthropic launches Claude Duet, an AI agent built on Claude Opus 5.5
Anthropic has announced Claude Duet, an AI agent aimed at professionals and small business owners. The Verge describes it as a competitor to Meta's Muse and OpenAI's Dots. It launches as a research preview, according to Anthropic, and is built around Claude Opus 5.5. It is available as a standalone app on iOS and Android and through Claude on the web.
Anthropic describes the daily routine this way. Each morning Duet sends a short brief on the most important items across the user's connected apps. Through the day it keeps tabs on those apps and messages the user only when something needs attention. Users can also hand it tasks of all shapes, from simple administrative jobs such as scheduling meetings to more complex ones: building dashboards with dozens of tools, creating financial or sales forecasts, prototyping, and building web apps.
Under the hood, The Verge reports, Duet pulls in memory, skills and connected apps from Claude, and it has a virtual machine with a browser that it uses to complete work. Before any big step, such as sending a payment or deleting something, it messages the user for approval.
Because Duet may act on sensitive information, it ships with a new dedicated classifier and safety system prompt. Anthropic says these are designed to reject requests to share the strictest tier of information, with credit card numbers given as an example.
Anthropic is open about the limits. It calls Duet an early version that still might misread a thread or miscalibrate what's urgent. It is also not as strong at personal errands such as booking tickets.
To try it, users can get on a waitlist, and they need to be a Pro or Max subscriber in the US.
Key facts
- Claude Duet is Anthropic's new agent for professionals and small business owners, launching as a research preview and built around Claude Opus 5.5.
- Each morning it sends a short brief on key items across connected apps, and during the day it messages the user only when something needs attention.
- It can handle tasks from scheduling meetings to building dashboards with dozens of tools, financial or sales forecasts, prototyping and web apps, using a virtual machine with a browser.
- It asks for approval before big steps like sending a payment or deleting something, and a new classifier and safety system prompt are meant to block requests for the strictest tier of information, such as credit card numbers.
- Access is through a waitlist, and users must be Pro or Max subscribers in the US.
Why it matters
Claude Duet puts Anthropic into a category The Verge frames as a race with Meta's Muse and OpenAI's Dots: agents that work on a user's behalf across their apps, not just answer questions. The pitch is proactive work. A daily brief and selective nudges replace a chat window the user has to open, and the agent can also take on larger jobs such as dashboards, forecasts and web apps.
Who it affects
Anthropic positions Duet for professionals and small business owners. In practice, the people who can try it first are Pro or Max subscribers in the US who get off the waitlist. Anyone whose work runs through connected apps, meetings and routine admin is the target user described.
How to use it
Get on the waitlist; a Pro or Max subscription in the US is required. The Verge says Duet is available as a standalone app on iOS and Android and through Claude on the web. Once in, the user connects apps, receives a morning brief, and can hand over tasks. Duet messages the user before big steps such as payments or deletions, so approvals arrive as messages. No price beyond the Pro or Max requirement is given.
How solid is it
The account is based on Anthropic's own announcement as relayed by The Verge, so the capabilities are the company's description, not independent testing. No benchmarks or performance numbers are given. The Verge describes Duet as available on iOS, Android and the web while also saying a waitlist applies; the source does not reconcile the two. No details about Muse or Dots are given beyond their being named as competitors.
Risks and caveats
Anthropic itself calls Duet an early version that still might misread a thread or miscalibrate what's urgent, and says it is not as strong at personal errands like booking tickets. Because the agent may act on sensitive information, Anthropic added a new dedicated classifier and safety system prompt meant to reject requests to share the strictest tier of information. The source does not explain how the classifier works or what that tier covers beyond credit card numbers. Approval prompts before payments and deletions are a safeguard, but the agent still decides what counts as urgent.
“still might misread a thread or miscalibrate what’s urgent.”
— Anthropic, describing Claude Duet's current limits, as quoted by The Verge