Anthropic Launches Claude Duet: The AI Agent That Works Before You Ask

 

Anthropic Launches Claude Duet: The AI Agent That Works Before You Ask

Most AI assistants wait for you to type. Anthropic's newest one messages you first. On October 21, 2026, the AI lab opened a research preview of Claude Duet, a standalone app in which Claude works as a proactive agent for professionals and small-business owners — sending a morning brief of what needs attention, checking connected apps through the day, and doing real work on its own cloud computer while you step away. It is the company's answer to OpenAI's always-on Dots agents and Meta's Muse agent, and it arrives with one detail every heavy user should read twice: Duet's background activity counts against the same usage limits as the rest of your Claude plan, which Anthropic says could push users to those limits "sooner than with chat alone."

Claude Duet is a standalone app for iOS, Android and the web — built around Claude Opus 5.5 — that messages you first instead of waiting for a prompt. (Photo: Sam Lion / Pexels)

Key Facts

• What: Claude Duet — a research-preview app in which Claude works as a proactive AI agent that comes to you with what needs attention, completes tasks you ask of it, keeps working when you step away, and checks with you before it acts on your behalf.
• When: Announced October 21, 2026; access granted via waitlist, with invitations going out in waves.
• Who gets it: US subscribers on Claude Pro and Max plans only. Not available on Team or Enterprise plans yet — Anthropic says it plans to expand there later.
• The model: Built around Claude Opus 5.5, Anthropic's flagship model for complex, long-horizon agentic work.
• The price catch: Included at no additional cost, but shares the plan's existing usage limits — background checks included. More activity = faster limit burn.
• The safety contract: Duet shows exactly what it is about to do and waits for approval before sending a message, making a payment, signing a document or deleting anything. Trading securities and moving money between accounts are off-limits even with approval.

What Claude Duet Actually Is

Claude Duet flips the oldest assumption in chatbots: instead of waiting for a prompt, it decides for itself when to talk to you. In Anthropic's own words, it is "a research preview of Claude that you message the way you'd message a person. Tell it what you need, typed or spoken, and it works in the apps you've connected, lets you know when something needs you, and is designed to ask before it sends a message, makes a payment, signs a document, or deletes anything."

The company says it designed Duet for "professionals and small business owners with more on their plates than they can track" — the people most likely to find an always-watching assistant useful rather than creepy. There is no extra onboarding step: Duet carries over the memory, skills and connected apps from your existing Claude account, can even search past conversations so you don't have to repeat yourself, and you talk to it in one ongoing thread. Separate channels can be started for different topics, and they share the same memory.

One interaction detail: Duet has a Tasks tab for tracking ongoing work and a Make tab where finished output lands as reviewable files — and only you can mark a task as finished. The agent does the work; the human closes it.

How It Works: The Morning Brief, Then Quiet

The daily rhythm is the product: one brief in the morning, then silence unless something matters. Each morning Duet looks across your email and calendar, picks what most needs your attention, and can take the first steps — drafting replies for you to review, pulling a brief together before a meeting. Through the day it checks your connected accounts every few hours and messages you only when something needs attention, then sends updates as work progresses and follows up on threads that have gone quiet.

That "every few hours" cadence is deliberate, and it connects directly to the usage-limits catch below: Anthropic says checking nonstop would use up a plan's limits much faster. Users can tell Duet which updates they want and how often, but the company has effectively designed the politeness of the agent around the economics of inference.

For sites and services without a connector, Duet falls back to its own browser in the cloud. It also supports plugins — Anthropic specifically mentions a small-business plugin — and can suggest ones that fit your work.

The Cloud Computer: Work Continues When You Close the App

Duet does not need your laptop or phone to stay on, because the work does not happen on your devices. The agent runs on its own cloud-based computer — a virtual machine with its own browser, complete with memory, skills and connected apps — so a week-long project can keep moving while your phone sits in your pocket. When a task finishes or needs your review, Duet sends a notification and an update in the main conversation thread.

The task range runs from the mundane to the ambitious. On the simple end: scheduling meetings, drafting replies and follow-ups, chasing missed Gmail and Slack messages. On the complex end: building dashboards, financial or sales forecasts, web apps and prototypes, preparing payroll for review, setting up a Shopify store, sending DocuSign agreements, and checking QuickBooks or Stripe for unpaid invoices.

One honest limitation comes with that cloud computer: in this early version, you cannot yet watch Duet's own browser as it works. The tasks appear; the browser stays behind the curtain.

Duet plugs into more than 6,000 connectors and plugins in the Claude Marketplace — Gmail, Slack, QuickBooks, Stripe, Gusto, HubSpot, DocuSign, Shopify and more. (Photo: Greta Hoffman / Pexels)

The Approval Model: What It Can and Cannot Do Without You

This is the section to read if you run a business: Duet reads, drafts and prepares work on its own, but the consequential actions wait for you. Before it sends a message, makes a payment, signs a document or deletes anything, Anthropic says, it "shows you exactly what it is about to do and waits for your approval." Payments go through Stripe's Link wallet: Duet confirms the item, Link asks you to approve the spend, and it pays with a virtual card so your real card number is never shared — Stripe says Link already powers payments for more than 200 personal agents, including Meta's Muse. Shop Pay from Shopify is also supported.

Beyond the approval queue, Anthropic drew two hard lines. Trading securities and moving money between accounts are off-limits even with approval — the agent simply will not do them. And a dedicated classifier plus a safety system prompt check each message for sensitive-information leaks: the classifier is designed to reject requests to share credit card numbers and government identification, and things like health or personal finances go out only when you have explicitly told Duet to send them. Anthropic also says Duet never sees your passwords or logins when you grant it access to accounts.

The early tester Anthropic chose to highlight tells the intended story: Amin Ebrahimi, who runs Shahs of Kabob in South Florida with roughly 110 employees, has used Duet for accounting, customer support and marketing. "I don't need to open up 100 tabs," he said, according to Anthropic. That is the pitch — not a smarter chat window, but a chief of staff that closes tabs for you.

Availability and the Usage-Limit Reality Check

Here is the catch, stated plainly because it changes the math for everyone on the waitlist. Claude Duet is included in existing Pro and Max subscriptions at no extra charge — but it shares the plan's existing usage limits with the rest of Claude, and its background checks count toward them. Anthropic warns this could cause users to reach their limits "sooner than with chat alone," and larger jobs like building a financial model or a deck take even more model turns. The every-few-hours checking cadence is the company's own answer to this: nonstop monitoring would burn through a plan fast.

Access is waitlist-only, starting with US Pro and Max subscribers, with invitations going out in waves. Team and Enterprise plans are not included yet, though Anthropic says it plans to expand there later.

The Proactive-Agent Race: OpenAI Dots, Meta Muse, and Now Duet

Claude Duet is the third major entry in a race that started this autumn: agents that reach out before you ask. Meta pushed the idea into the spotlight with its Muse AI agent, which drew millions of downloads and investor attention toward autonomous AI services. OpenAI followed with always-on "Dots" agents that pursue user goals across apps like Slack and Microsoft Teams. Anthropic is now the late but heavyweight entrant — and it is pitching trust as its differentiator, building Duet on what it calls its framework for trustworthy agents, the same framework it says guides Claude Code and Claude in Chrome.

The aesthetic contrast is almost a summary of each company's strategy. OpenAI's Dots launched with friendly mascots; Meta's Muse is a Totoro-like creature; Duet's identity is a scribble, in line with every other Anthropic product. The company is aiming at professionals doing consequential work, not consumers who want a friendly face.

Anthropic acknowledges Duet "still might misread a thread or miscalibrate what's urgent." Early reviewer Dan Shipper of Every found its email drafts weaker than a rival agent's and noted its voice capability is currently "just transcription" rather than a real voice mode.

Who Should Actually Join the Waitlist

The honest answer: US-based Pro or Max subscribers who live in connected apps and hate tab-juggling. If your day is a sequence of inbox, calendar, invoices and follow-ups, Duet's morning brief plus background task handling is exactly the relief Anthropic is selling. If you are on a Team plan, outside the US, or want to watch it work in its browser before trusting it with money-adjacent tasks, wait for the preview to mature.

FAQ

What is Claude Duet?
A research-preview app, announced by Anthropic on October 21, 2026, in which Claude works as a proactive agent. It sends a morning brief from your connected apps, checks them every few hours, completes assigned tasks in the background on its own cloud computer, and asks approval before sending messages, making payments, signing documents or deleting anything.

How do I get Claude Duet?
Join the waitlist. It is currently limited to US subscribers on Claude Pro and Max plans, with invitations going out in waves. It runs as standalone iOS and Android apps and on claude.ai. Team and Enterprise plans are not supported yet.

Does Claude Duet cost extra?
No — but that is not the whole story. It is included in Pro and Max subscriptions and shares the plan's existing usage limits with the rest of Claude. Its background checks count toward those limits, and Anthropic says users may hit limits "sooner than with chat alone." Bigger tasks take more model turns.

What can Claude Duet actually do?
Morning briefs from email and calendar, meeting scheduling, drafting replies and follow-ups, chasing unpaid invoices (QuickBooks, Stripe), preparing payroll (Gusto), building dashboards, forecasts, web apps and prototypes, setting up a Shopify store, sending DocuSign agreements, and buying supplies via Stripe Link or Shop Pay — always with approval for consequential actions.

Is it safe to give Claude Duet access to my accounts?
Anthropic's design: Duet never sees your passwords or logins; a dedicated classifier checks each message for sensitive-information leaks; it shows you exactly what it will do and waits for approval before sending, paying, signing or deleting; trading securities and moving money between accounts are blocked even with approval. Anthropic still cautions the agent may misread a thread or misjudge urgency.

How is this different from OpenAI Dots or Meta Muse?
Same category — proactive agents that work on behalf across apps. OpenAI's Dots and Meta's Muse launched earlier and target broader audiences; Duet is Anthropic's play for professionals and small businesses, gated as a US-only research preview for Pro/Max subscribers.

Sources

Sources: Anthropic announcement via 9to5Mac (quoting Anthropic verbatim), "Anthropic takes on ChatGPT Dots and Meta Muse with Claude Duet," October 21, 2026 — 9to5mac.com; Reuters, "Anthropic unveils Claude Duet AI agent app," October 21, 2026 — reuters.com; MacRumors, "Anthropic Launches Claude Duet, a Proactive AI Agent," October 21, 2026 — macrumors.com; WebProNews, "Anthropic Opens Claude Duet Preview," October 21, 2026 — webpronews.com; Engadget, "Claude Duet Is Anthropic's Answer To Meta Muse And OpenAI Dots," October 21, 2026 — engadget.com; SiliconANGLE, "Anthropic introduces Claude Duet, a proactive AI agent for professionals," October 22, 2026 — siliconangle.com; PYMNTS, "Anthropic's New AI Agent Purchases Business Supplies," October 21, 2026 — pymnts.com; Unite.AI, "Anthropic Opens Claude Duet Research Preview to US Pro and Max Users," October 21, 2026 — unite.ai. Verbatim Anthropic quotes in this article are from the company's October 21, 2026 announcement as quoted by 9to5Mac and the official @claudeai account. Images: Pexels (free license) — hero by Sam Lion; workspace photo by Greta Hoffman.