In 20 seconds
- Anthropic introduced Claude Duet on October 21, 2026: a proactive agent in its own iOS, Android and web app that sends a morning brief, checks connected accounts every few hours and runs tasks on a cloud virtual machine. It runs on Claude Opus 5.5.
- The research preview is open by waitlist to Claude Pro and Max subscribers in the US only. Team and Enterprise plans aren't included yet. It costs nothing extra, but its work, including background check-ins, counts toward your plan's five-hour and weekly limits.
- Duet asks for approval before it sends a message, makes a payment, signs a document or deletes anything. Anthropic says trading securities and moving money between accounts are off-limits even with approval.
On this page
Claude Duet is a proactive agent from Anthropic that lives in its own app and messages you first. Anthropic opened it on October 21, 2026, as a research preview, by waitlist, for Claude Pro and Max subscribers in the US. Unlike a chat, which waits for your question, Duet sends a morning brief, checks your connected accounts every few hours and works on tasks in the background. It asks for your approval before it sends, pays, signs or deletes anything. It arrives three weeks after OpenAI launched dots, its always-on ChatGPT agents, and six weeks after Meta's Muse agent app.
This explainer is based on Anthropic's announcement, the Claude Duet product page and two Claude help center articles, checked October 23, 2026, with reporting from 9to5Mac, 9to5Google and MacRumors.
The short version
- What: a personal agent you message by text or voice, in one ongoing thread. It runs on Claude Opus 5.5, on a cloud virtual machine with its own browser, so your phone or laptop doesn't need to be on.
- Who: US Pro and Max subscribers, admitted from a waitlist in waves. Team and Enterprise aren't included yet.
- Where: standalone iOS and Android apps, and the web at claude.ai. It's separate from the main Claude app.
- Cost: no extra charge during the preview, but Duet's work, including background checks, counts toward your plan's usage limits.
- Approvals: it waits for you 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 is Claude Duet?
Claude Duet is Anthropic's always-on assistant for "professionals and small business owners," as the announcement puts it. Each morning it sends a short brief on the most important items across your connected apps. Through the day it checks those accounts and contacts you only when something needs you. Anthropic's article, "Introducing Claude Duet," says it "checks your connected accounts every few hours."
You can also hand it work. Anthropic's examples include scheduling meetings, finding missed Gmail and Slack messages, checking QuickBooks or Stripe for unpaid invoices and drafting reminders, preparing payroll in Gusto, routing agreements through Docusign, building dashboards and forecasts, and prototyping web apps. Larger jobs come back as editable files in a Make tab, and ongoing work appears in a Tasks tab, according to the help center. Duet may suggest a task is done, but only you can mark it complete.
Duet carries over your Claude memory, skills and connectors from your first message. Anthropic says it can use more than 6,000 connectors and plugins from the Claude Marketplace, and a cloud browser for sites without a connector.
How do I get Claude Duet and join the waitlist?
You join from the Claude Duet product page. The help center lists the steps:
- Choose Join the waitlist and enter your name and email. Use the email on your Claude account.
- Pick your plan and your phone type (iPhone or Android), then submit.
- Wait for an invitation email. Access goes out in waves, and joining doesn't guarantee a date.
- Open the invitation on your phone and sign in with the Apple or Google account tied to the email you registered. The help center says it "can take a few minutes for the app to become available."
Anthropic hasn't given a date for wider release.
Which plans get Claude Duet, and is it available in my country?
Only Claude Pro and Max in the United States, for now. Anthropic's announcement says the preview is "available by waitlist today, starting with US Pro and Max subscribers." The help center says Duet "isn't available on Team or Enterprise plans yet," and that those plans are planned for later. Paying with Link also needs a US payment method. Anthropic hasn't named other countries or dates.
| Item | Claude Duet research preview |
|---|---|
| Announced | October 21, 2026 |
| Plans | Claude Pro and Max (Team and Enterprise not yet) |
| Countries | United States only |
| Access | Waitlist, invitations in waves |
| Apps | iOS, Android, web at claude.ai |
| Model | Claude Opus 5.5 (no model choice) |
| Extra price | None during the preview |
| Usage | Shares your plan's five-hour and weekly limits, including background work |
| One-time bonus | One full limit reset for admitted users, available until November 30, 2026 |
Table 1Duet is free to try for US Pro and Max subscribers, but its background work uses the same limits as chat. Source: Anthropic's announcement, product page and "Get started with Claude Duet" help article.
Last verified: October 23, 2026
The usage point matters. The help center says Duet draws on your limits "at the same rate as regular chat," and that background work such as the morning brief and check-ins also counts, so "you may hit limits sooner." For how those limits work, see our Claude app explainer.
What does Claude Duet ask permission for?
Anthropic splits Duet's actions into three groups. It reads, drafts and prepares work on its own. Before a consequential action, the announcement says, "it shows you exactly what it's about to do and waits for your approval." If you can't approve right away, the help center says, it brings the request back later. A few actions are blocked outright.
Here is how Anthropic's examples fall into those groups:
| Duet does on its own | Duet asks first | Off-limits, even with approval |
|---|---|---|
| Reads connected apps and inboxes | Sending a message or email | Trading securities |
| Drafts replies, briefs, decks and models | Making a payment (Link or Shop Pay) | Moving money between accounts |
| Checks accounts every few hours | Signing a document | |
| Builds dashboards and prototypes | Deleting anything |
Table 2Duet acts alone only on reading and drafting; anything that leaves your account waits for you. Source: Anthropic's announcement and help center, October 21–22, 2026.
Purchases have a second step. With Link by Stripe, the help center says Duet "will not submit the Link payment until you approve the maximum amount" in the Link app, and each purchase uses a one-time card scoped to that merchant and amount. Duet sees your saved cards' type, last four digits and expiration date, but not the full card number. Recurring payments aren't supported.
What are Claude Duet's privacy and data controls?
Anthropic describes these controls:
- Credentials: Duet "never directly sees" your passwords or sign-in tokens. Passwords go into a separate form that passes them to the browser, outside the chat.
- Sensitive data: a dedicated classifier is designed to "reject requests to share the strictest tier of information, including credit card numbers and government identification." Health and personal finance details are sent only when you've explicitly told Duet to.
- Connections: you start every connection yourself and can disconnect any app. Disconnecting stops future access, but Duet keeps memory of what it already learned.
- Past chats: Duet can search your Duet chats and your non-project Claude chats. You can turn that off with Search and reference chats under Settings > Memory.
The help center also warns that Duet treats emails and web pages as information, not commands, but that no defense against prompt injection is perfect. Our prompt injection explainer covers why that risk is hard to remove for agents that read your inbox. Anthropic says more safety details will follow "in the coming days." Its pages checked for this story don't describe data retention periods or whether Duet data is used for training beyond your existing Claude settings.
What can't Claude Duet do yet?
Anthropic lists these limits for the preview:
- Microsoft 365 "isn't available yet," and Slack works with limits.
- It can't make phone calls or send text messages.
- You can't watch its cloud browser or take control of it.
- It can't reach local files or all of your Claude projects, and it can read but not edit your Claude work.
- It "still might misread a thread or miscalibrate what's urgent," and can miss something that arrives between check-ins. Anthropic also calls personal errands such as booking tickets a weaker area.
How is Claude Duet different from Claude, OpenAI's dots and Meta's Muse?
Inside Claude, Duet is the proactive layer. The Claude app and Cowork wait for you to start a task. Duet watches your accounts, starts conversations and runs on its own virtual machine.
The closest rivals are OpenAI's dots and Meta's Muse. 9to5Mac and 9to5Google framed Duet as an answer to both.
| Claude Duet | OpenAI dots | Meta Muse | |
|---|---|---|---|
| Launched | October 21, 2026 (preview, waitlist) | September 29, 2026 | September 8, 2026 (US) |
| Model | Claude Opus 5.5 | GPT-6 Astra | Muse Spark |
| Who | US Pro and Max | ChatGPT Pro and Business Premium; excludes EEA, UK, Switzerland | US users, free; $20 and $100 plans for heavier use |
| Runs on | Cloud VM with its own browser | Cloud computer and browser; laptop opt-in | Cloud VM |
| Approvals | Asks before sending, paying, signing, deleting | Custom rules: act, act if pre-approved, ask first, hand off | Asks before sensitive actions such as email or purchases |
Table 3All three ask before consequential actions, but they differ on plans, regions and how approvals are set. Source: Anthropic (October 21, 2026); OpenAI's dots documentation as reported in our October 12 explainer; Meta and TechCrunch as reported in our Muse Spark explainer.
Last verified: October 23, 2026 (Duet); dots and Muse details as of our earlier stories
Two differences stand out. OpenAI's dots let you open the agent's cloud computer and click "Take over"; Duet's preview doesn't let you watch its browser. And plan access differs: dots are included with Pro (from $100 a month) and Business Premium, while Duet is included with Claude Pro ($20 a month) and Max. Duet is US-only and waitlisted. Google's Gemini agent for work is a fourth, enterprise-only take on the same idea.
What's confirmed, reported and uncertain
- Confirmed (Anthropic): the October 21 preview; US Pro and Max by waitlist; iOS, Android and web; Opus 5.5; shared usage limits; the approval rules and off-limits actions; the classifier; the listed limitations; the limit reset until November 30, 2026.
- Reported: comparisons with dots and Muse come from press framing (9to5Mac, 9to5Google), not from Anthropic.
- Uncertain: when Duet opens to more people, other countries, or Team and Enterprise; data retention; the further safety details Anthropic has promised.
About this story
Based on the sources linked below. How we report: Editorial standards.
Sources · 10
- Introducing Claude Duet — Anthropic (October 21, 2026)
- Claude Duet product page — Anthropic
- Get started with Claude Duet — Claude Help Center
- How to use Link by Stripe with Claude Duet — Claude Help Center
- Introducing dots — OpenAI (September 29, 2026)
- Introducing Muse, a personal AI agent — Meta Newsroom (September 8, 2026)
Show all 10 sources
- Anthropic takes on ChatGPT Dots and Meta Muse with Claude Duet — 9to5Mac (October 21, 2026)
- Claude Duet is Anthropic's proactive AI agent answer to Muse and OpenAI's Dots — 9to5Google (October 21, 2026)
- Anthropic Launches Claude Duet, a Proactive AI Agent — MacRumors (October 21, 2026)
- Hacker News discussion of Claude Duet (Tier 3, October 21, 2026)






Comments (0)