What is Claude Duet and how to try it

On October 21, 2026, Anthropic launched Claude Duet as a research preview: a new app where Claude stops waiting for you to write and starts writing to you. It sends you a brief every morning, checks your apps every few hours, takes on big tasks in the background and asks for your OK before sending, paying or signing anything.

Here is what it is, what it really does, who it is for, what it costs and how to get it. Everything comes from Anthropic's official pages, linked at the end. And the most important point if you are in Latin America: for now it is only available in the United States.

What Claude Duet is

Claude Duet is a proactive agent: an assistant you hand work to by message, which works in the cloud with the apps you connect and writes to you when something is done or when it needs a decision from you. It runs on Claude Opus 5.5, and you cannot switch models.

The difference from the Claude you know is simple: Claude waits for you to open a chat. Claude Duet comes to you. It is a separate app from the Claude app, for iPhone and Android, and it also works from claude.ai on a computer.

Comparison between Claude and Claude Duet: Claude waits for you to open a chat; Claude Duet writes to you first, keeps working on its own cloud computer and checks your apps every few hours — Claude vs Claude Duet: the difference is who writes first.

What it does

A morning brief. It reviews your email and calendar, tells you what is a priority and can take the first steps, such as drafting replies.
It watches your apps. It checks your connected accounts every few hours and only writes when something needs you. You decide what you want to hear about and how often.
Big tasks in the background. Scheduling meetings, building a dashboard from several tools, financial or sales forecasts, prototypes and web apps. Larger jobs come back as files (docs, sheets or decks) in the Make tab.
Its own cloud computer. It works on a virtual machine with a browser, so it keeps going even if you close the app or turn off your phone. For sites without a connector, it uses that browser.
The Claude you already use. From the first message it has your memory, your skills and your connected apps from Claude, all in one thread.
Purchases with your OK. It can buy with Stripe's Link, using a one-time virtual card, or with Shopify's Shop Pay. Both are set up inside Claude Duet.

According to Anthropic, it works with more than 6,000 connectors and plugins. Official examples include Gmail, Google Calendar, Google Drive, Slack, QuickBooks, HubSpot, Docusign, Calendly, Gusto, Notion, Canva, Shopify and Stripe. Microsoft 365 is not available yet.

A day with Claude Duet in five steps: it sends your brief in the morning, checks your apps during the day, takes on big tasks when you ask, asks for your OK before acting and hands you files when done — How it works in the background, based on the official description.

The cases Anthropic shows are real work: finding unpaid invoices in QuickBooks or Stripe and drafting the reminder, preparing payroll in Gusto, getting a contract ready to sign in Docusign or drafting follow-ups with a deal's history in HubSpot. One of the people who tried it before launch said Duet noticed a client call clashing with her flight and offered to move it.

Who it is for

Anthropic built it for professionals and small business owners with more on their plate than they can track. Its examples are a real estate broker, a restaurant owner and a brand strategist. The demo on the official page follows a café owner opening a new location.

It is not meant for personal errands: Anthropic acknowledges it performs worse on tasks like buying tickets than on work.

Safety: what it does alone and what it asks

This is the part I like most, because it decides whether an agent like this can be used in a business. Claude Duet reads, drafts and prepares on its own. But before sending a message, paying, signing a document or deleting anything, it shows you the exact action (to whom, what and how much) and waits for your approval.

Claude Duet safety in three columns: what it does alone, what it asks first, such as sending messages, paying, signing or deleting, and what it never does, such as trading securities or moving money between accounts — What it does alone, what it asks and what it never does.

Never, even with approval: trading stocks or securities, moving money between accounts or accessing someone else's account.
Your passwords: Claude Duet never sees passwords, cookies or sign-in tokens. You type them into a form in the app and they reach the browser encrypted.
A second check outside the model: each action is reviewed by a separate classifier to confirm it matches what you asked. That way, instructions hidden in a web page cannot change the outcome.
Sensitive data: another classifier checks every message before it goes out. It never shares credentials, full card or account numbers or ID documents. Health or finance information only goes out if you ask.
Testing: it went through internal red teams, a penetration test by an outside firm, and it is covered by Anthropic's public bug bounty.
Your history: it keeps the conversation and memory until you delete or reset them. Disconnecting an app cuts access from that moment, but what it already learned stays until you reset its memory.

One detail the safety page does not spell out: whether that data is used to train models. If that matters for your business, check the privacy settings of your Claude account.

What it does not do well yet

Anthropic is upfront about the limits, and they are worth reading before you hand it anything important:

It can misread a thread or misjudge how urgent something is.
Since it checks in at intervals, something urgent can arrive between one check and the next.
It does not make phone calls or send texts. With your OK, it does send emails from your connected accounts.
You cannot watch or take over its browser yet.
It cannot see your local files, and it can read your Claude work but not change it.
Anthropic itself warns that asking for approval so often gets tiring, which can lead to approving without reviewing.

Dan Shipper at Every used it for about two weeks before launch. He called it the first personal agent he would consider a pro tool, though he found its email drafts weaker than those of Dot, OpenAI's agent, and that scheduled tasks cannot reach custom connectors.

What it costs

During the preview, Claude Duet is included in the Pro and Max plans at no extra cost. It has no separate price, but it uses the same usage limits as your plan: the 5-hour one and the weekly one. Background work, like the morning brief, also counts, so you can hit your limit sooner than with regular chat.

To make up for it, people who get in receive one full reset of both limits, available until November 30, 2026.

How to get it, step by step

You need three things: a Pro or Max plan, a Claude account in the United States and an invitation from the waitlist. Team and Enterprise plans do not have it yet.

Go to the official Claude Duet page on claude.com and choose "Join the waitlist".
Enter your name and email. Use the same email as your Claude account, because access is tied to that email.
Choose your plan and phone type (iPhone or Android). Optionally, say what you would hand off first.
Submit and wait for the email. Invitations go out in waves, so joining does not guarantee a date.
When it arrives, open it on your phone, follow the iPhone or Android steps and sign in with your Claude account. The app can take a few minutes to activate.

What about Panama or Latin America?

Not yet. Anthropic's help center is clear: during the preview, Claude Duet is only available in the United States. I would not try workarounds like a VPN: access is tied to your account, your plan and your email, and Anthropic limited it on purpose while it tunes the safety side.

What you can do now is get ready. Claude Duet inherits your memory, skills and connectors from Claude, so everything you set up today in the Claude app (Gmail, Calendar, Drive, Slack and the rest of your business tools) will be ready from the first message the day it reaches your country.

I will test it as soon as it is available in Panama and update this article with the real test: what I handed it, what it did well and what it did not.

My take

Claude Duet is not another chat. It is the step from "I ask the AI" to "the AI works for me and checks with me". For a business owner in the region, the value is in three things: nothing slips through between email, calendar and invoices; the heavy work (dashboards, forecasts, contracts) happens in the background; and nothing goes out in their name without their approval.

It is not alone: OpenAI has Dots and Meta has Muse in the same category. What sets Duet apart is how careful it is with permissions, and that it inherits everything you already have set up in Claude.

If you want something like this today

You do not have to wait for Duet to have AI working on its own in your business. My team builds assistants that alert you on WhatsApp or Telegram when a lead comes in, CRMs that follow up on their own and dashboards that update without anyone touching them. If you want to see it applied to your case, book a VIP Consulting session.

Book your VIP Consulting session → (the booking page is in Spanish)