Claude Duet is Anthropic's proactive agent app, in preview

Anthropic opened a research preview of Claude Duet on October 21, 2026, a separate app in which Claude acts as a proactive agent. Instead of waiting for a prompt, Duet watches your connected apps and messages you when something needs attention. It is a waitlist-only preview for Pro and Max subscribers in the US.

What Claude Duet does

An agent is an AI system that carries out tasks on its own, not just answers questions. Duet starts each day with a morning brief of the top items across your connected apps, according to Anthropic's announcement. It then checks those accounts every few hours and messages you only when something needs you.

The work happens on Duet's own virtual machine, a cloud computer with its own web browser. Your laptop or phone does not need to stay on. Anthropic says tasks range from scheduling meetings to building dashboards, forecasts, prototypes and web apps.

Duet carries over the memory, skills and connected apps from your regular Claude account. It is designed around Claude Opus 5.5.

Detail	What Anthropic says
Access	Waitlist, starting with US Pro and Max subscribers
Price	No separate charge; it shares your plan's usage limits
Apps	iOS and Android apps, plus claude.ai on a computer
Not yet	Team and Enterprise plans
Connectors	Over 6,000, including Gmail, Slack, QuickBooks, Stripe and Shopify

MacRumors lists only the web and an iOS app, while Anthropic's pages and 9to5Mac also name Android.

How it handles money and mistakes

Duet asks before it acts on anything that matters. Anthropic says it waits for approval before it sends a message, makes a payment, signs a document or deletes anything. Trading securities and moving money between accounts are off-limits even with approval.

Purchases run through Link, Stripe's digital wallet, and Shop Pay checkout. 9to5Mac reports that Duet can also use Stripe to check unpaid invoices and draft payment reminders. Only you can mark a task as finished.

Anthropic says Duet never handles passwords or sign-in tokens directly, keeping them apart as its Managed Agents product does. A classifier and a safety system prompt screen each message for sensitive data. The most sensitive kind, such as credit card numbers and government IDs, is rejected outright.

The company is open about the limits. Duet "may misread a thread or miscalibrate what's urgent," its product page says. Personal errands such as booking tickets are less developed, and Duet cannot yet show you its browser.

Where it fits

Duet is Anthropic's answer to two rivals, MacRumors notes: Meta's Muse agent and OpenAI's Dots. Tech AI Wire covered that contest in OpenAI launches Dots, an always-on rival to Meta's Muse.

Anthropic's examples come from small businesses. Jess Provencher, a broker who owns Pro Homes, said: "[This is] exactly the kind of thing I want caught without having to ask." Amin Ebrahimi, owner of Shahs of Kabob, put it more simply: "I don't need to open up 100 tabs."

What this means for developers

Duet is sold to individual subscribers for now, but it shows where agent products are heading, and that shapes what you build.

• Expect usage to climb quietly. Background checks count toward plan limits, and larger tasks use more model turns. Teams trying Duet on paid plans should watch how fast limits run down.
• Make your app easy for an agent to reach. Duet works through connectors and falls back to a cloud browser for sites without one. A clean connector or API means an agent can use your product without scraping it.
• Design approval steps into risky actions. Anthropic's line of approve-first, with money transfers banned outright, matches how banks are responding, as in Schwab, JPMorgan block AI agents from moving client money. Build the same checkpoints into agent features of your own.
• Watch the Team and Enterprise rollout. Duet is not on those plans yet. When it arrives, admins will need rules for which connected apps an agent may touch.