Anthropic Claude Duet, the company’s new AI agent for professionals and small businesses, entered research preview on October 21 with a waitlist for eligible subscribers. Anthropic says the product is meant to watch connected workplace apps, flag items that need attention and carry out requested work, a more active role than a chat window waiting for the next prompt.
The company is positioning Duet against the emerging agent products from Meta and OpenAI, respectively called Muse and Dots. That is a category comparison, not evidence that the products perform the same tasks or to the same standard. Anthropic’s stated focus is work, and it says the early product is less capable at personal errands such as booking tickets.
What can Anthropic Claude Duet do?
According to Anthropic, Duet can produce a morning briefing based on connected apps and continue monitoring those apps during the day. The company says it can notify a user when something requires attention, rather than sending an alert for every new item.
Anthropic also says users can assign tasks ranging from scheduling meetings to building dashboards across many tools, creating financial or sales forecasts, prototyping and building web apps. Those are company-stated capabilities, not independently tested results.
Duet draws on a user’s Claude memory, skills and connected apps, according to reporting by The Verge. It also has access to a virtual machine with a browser to complete work. In practical terms, that setup is intended to let the agent work through a task across software services over time, rather than only produce text in a conversation.
What actions need approval?
Anthropic says Duet must ask before consequential steps, including sending a payment or deleting something. The company also says it added a dedicated classifier and safety system prompt designed to reject requests involving its strictest category of sensitive information, with credit-card numbers cited as an example.
Those controls do not amount to a guarantee that the agent will interpret workplace context correctly. Anthropic says the preview version can misread a thread or judge the urgency of an item incorrectly. That warning is particularly relevant for a tool whose pitch involves deciding when to interrupt users and taking actions in connected accounts.
Who can use Claude Duet?
Anthropic has opened a waitlist rather than broadly releasing the product. The Verge reported that access in the United States requires a Pro or Max subscription; CNBC likewise reported that Pro and Max subscribers are eligible during the preview. Duet is available through standalone iOS and Android apps as well as Claude on the web, according to The Verge.
Anthropic’s preview arrives as Meta’s Muse and OpenAI’s Dots seek users in the same broad agent market. For now, the useful distinction is narrower: Duet is a gated workplace-focused preview, and Anthropic itself says it remains an early version.
This story draws on original reporting from The Verge.
