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

Anthropic opened a research preview of Claude Duet on 21 October 2026, a standalone app in which Claude works as a proactive agent [1][4]. It sends a morning brief, takes on tasks by message and asks before acting. Access runs through a waitlist for Pro and Max subscribers in the US [1][2].

What does Claude Duet do?

Anthropic says Claude Duet is built for professionals and small business owners with more on their plates than they can track [1]. Each morning it sends a short brief on the most important items across connected apps, then messages users only when something needs their attention [1].

Users can also hand it larger tasks, from scheduling meetings to building dashboards, creating financial or sales forecasts, prototyping and building web apps, according to Anthropic [1]. It works on a virtual machine with its own browser, so the user's laptop or phone does not have to stay on [1].

Duet is built around Claude Opus 5.5 and carries over memory, skills and connected apps from the user's Claude account [1][4]. Anthropic says it can use more than 6,000 connectors and plugins from the Claude Marketplace [1]. The product FAQ says it can find unpaid invoices, prepare payroll for review and get contracts ready to send [2].

Who can use it, and what does it cost?

Anthropic opened a research preview of Claude Duet on 21 October 2026 for Pro and Max subscribers in the US. Team and Enterprise plans are not included yet, the product FAQ says [2].

There is no separate price. During the preview, Duet shares the plan's usage limits with the rest of Claude, and its background checks count toward them [1][2]. Duet is a standalone app for iOS and Android and can be opened on the web at claude.ai [2].

How does Anthropic handle permissions and safety?

Claude Duet asks for approval before it sends a message, makes a payment, signs a document or deletes anything. Some actions, such as trading securities or moving money between accounts, are off-limits even with approval, Anthropic says [1].

According to Anthropic, Duet never directly sees passwords or sign-in tokens, and a new classifier and safety system prompt check each message for leaks of sensitive information [1][4]. Purchases can run through Stripe's Link wallet with a virtual card [1].

Anthropic also lists limits: Duet may misread a thread or miscalibrate what's urgent, is weaker at personal errands such as booking tickets, and cannot yet show its browser [1]. It checks connected accounts every few hours rather than nonstop, to save plan limits [1].

How does Duet fit the agent race?

The Verge describes Claude Duet as a competitor to Meta's Muse and OpenAI's Dots [3]. Meta paused Muse's chat-sharing feature this week after researchers found more than 6,800 shared-chat links, as we reported on 21 October.

Our read

Opinion: Duet's pitch is less about raw capability than about restraint: approval gates, hard limits on money moves and a classifier for the most sensitive data. Whether that holds up will depend on how the preview behaves in real inboxes, which no outside tester has reported on yet.

What we could not verify

• Anthropic has not published a separate price; Duet draws on existing Pro and Max usage limits, but how quickly background checks use them up is not quantified.
• When Team and Enterprise plans, or users outside the US, will get access; Anthropic gives no date.
• How long the waitlist is and how many people are admitted in the first wave.
• Anthropic's claims about Duet's judgment and safety safeguards have not been independently tested; Anthropic says more safety details will follow.
• We found no separate post on anthropic.com/news; the announcement is on Anthropic's claude.com blog and the official @claudeai account.