Anthropic Launches Claude Duet, a Proactive Agent

Most AI assistants wait to be asked. Anthropic’s newest one is meant to speak up. The company has opened a research preview of Claude Duet, a standalone app in which Claude works as what Anthropic calls “a proactive agent that understands what needs your attention”. It sends a morning brief, takes on tasks by message, keeps working when you put your phone down, and checks with you before acting on your behalf.

Access is narrow for now. Claude Duet is on a waitlist for Claude Pro and Max subscribers in the US, and “Team and Enterprise plans are not included yet”, according to its product page. It runs as an app on iOS and Android and on the web at claude.ai, and Anthropic logged the launch in its release notes on 21 October. Readers in the UK will have to wait, as no timetable has been given for other countries.

What Anthropic says it does

Anthropic built Duet for “professionals and small business owners with more on their plates than they can track”, it says in its launch article. It runs on Claude Opus 5.5 and carries over a user’s existing memory, skills and connected apps from Claude, so there is little to set up.

Each morning it sends a short brief on the most important items across those apps, then, in Anthropic’s telling, “messages you only when something needs your attention”. It works on its own virtual machine with its own browser in the cloud, so your laptop does not need to stay on. Anthropic says it can handle tasks from booking meetings to building dashboards, financial or sales forecasts, prototypes and web apps.

It can draw on more than 6,000 connectors and plugins. Anthropic’s examples include chasing unpaid invoices in QuickBooks or Stripe, preparing payroll in Gusto, sending agreements through Docusign and setting up a Shopify store. It can also buy things through Link, Stripe’s digital wallet, which pays with a virtual card so the real card number is not shared.

Duet checks connected accounts every few hours rather than constantly, Anthropic says, because nonstop checking would use up a plan’s limits much faster. Those background checks count towards the same usage limits as the rest of Claude.

Approvals and limits

The safeguards take up much of the announcement. Before Duet sends a message, makes a payment, signs a document or deletes anything, Anthropic says it shows exactly what it is about to do and waits for approval. Some things are off-limits even with approval, such as trading securities or moving money between accounts.

Anthropic says Duet never directly sees passwords or sign-in tokens, that its actions are screened by the same classifier that guards Claude Code, and that red teams and an outside firm have tested it for security. A new classifier is meant to block it from sharing the most sensitive information, such as credit card numbers and government ID, while health or financial details go out only on explicit instruction. The company says it will share more about Duet’s safety design “in the coming days”.

It is also frank about the weak spots. “This is an early version of Claude Duet, and it still might misread a thread or miscalibrate what’s urgent,” Anthropic writes, adding that it can miss something that arrives between check-ins. Personal errands such as booking tickets “aren’t quite as strong”, and it cannot yet show users its browser.

A crowded field

The Verge frames Duet as a competitor to Meta’s Muse and OpenAI’s Dots, and notes that it is aimed squarely at professionals and small business owners. Anthropic’s pitch leans on judgement rather than reach. “What makes Claude Duet so helpful is its judgment,” the company says.

Its examples come from people it chose. A South Florida restaurant owner with about 110 staff says he has used it for accounting, customer support and marketing. A property broker says it caught a showing booked on the wrong day.

For now, Duet lives in a separate app from the main Claude app, and Anthropic says early users will shape how the two come together. Whether people want an assistant that messages them first is the question the waitlist is meant to answer.