Anthropic has launched Claude Duet, a proactive AI agent that monitors your connected apps, flags what needs your attention, and keeps working in the cloud after you close the app. It matters because the agent race is shifting from chatbots that wait for prompts to assistants that come to you first — and Duet is Anthropic’s direct answer to OpenAI’s Dots and Meta’s Muse.
October 21, 2026 — Anthropic announced Claude Duet on Wednesday as a research preview for professionals and small-business owners. The standalone agent is built around the company’s Claude Opus 5.5 model and is rolling out gradually, via a waitlist, to subscribers on the Pro and Max plans in the United States.
What Claude Duet actually does
Duet runs on its own cloud-based computer, complete with a browser, so it can keep working even when you have closed the app. It sends a short brief each morning covering important items from your connected apps, then checks those apps every few hours and alerts you when something needs attention. Anthropic says it can handle simple administrative work as well as larger jobs such as scheduling meetings, building financial forecasts, and building web apps.
Because Duet uses the same memories, skills, and connectors as a regular Claude account, setup is minimal. It can search your past conversations, and it connects to more than 6,000 connectors and plugins in the Claude Marketplace, including services such as Gmail and Slack. Work happens in one ongoing thread, tracked in a Tasks view, and only you can mark a task complete.
The guardrails: it asks before it acts
Anthropic is positioning human approval as a hard boundary. Duet is designed to ask before it sends a message, makes a payment, signs a document, or deletes anything. It integrates with Stripe to check for unpaid invoices and prepare payment reminders, and it can make purchases through Stripe’s Link wallet and Shopify’s Shop Pay — with your approval. Anthropic says Duet does not see passwords or logins when it accesses accounts, and a dedicated classifier and safety prompt check messages to guard against sensitive information leaking. Trading securities is off-limits entirely.
There are honest limits, too. Duet is included at no additional charge, but everything it does — including background checks — counts toward your existing plan usage limits, and Anthropic warns you may reach those limits sooner than with chat alone. The company also acknowledges the agent can misjudge urgency and is still less effective at personal tasks such as booking tickets. Team and Enterprise plans are not eligible yet, though Anthropic plans to add them later.
Why it matters
Every major AI lab now has an always-on agent: OpenAI launched its Dots agents last month, Meta’s Muse has drawn millions of downloads, and now Anthropic has Duet. The pattern to watch is the business model hiding underneath — a proactive agent that works all day will burn through usage allowances far faster than a chatbot you open twice a day, which quietly pushes heavy users toward pricier tiers. For subscribers, the practical test is simple: does the morning brief save you more time than the allowance it consumes? Everything AI, tested — that is the question we will keep asking as access opens up.
Frequently asked questions
Who can use Claude Duet right now?
Duet is a research preview rolling out via a waitlist to Claude Pro and Max subscribers in the United States. It is available through standalone iOS and Android apps and at claude.ai. Team and Enterprise plans are not included yet.
Does Claude Duet cost extra?
No. Anthropic says there is no additional charge, but all of Duet’s activity — including background work — counts toward your existing plan usage limits, so you may hit your cap sooner than with chat alone.
Can Claude Duet spend money on its own?
No. Duet is designed to ask for your approval before making a payment, sending a message, signing a document, or deleting anything. Purchases run through Stripe’s Link or Shopify’s Shop Pay with that approval step.
What model powers Claude Duet?
Duet is built around Anthropic’s Claude Opus 5.5 model, running on a cloud-based computer with its own browser.
Sources: Reuters, MacRumors, 9to5Mac, TestingCatalog