---
title: "Anthropic Launches Claude Duet, an AI Agent That Messages You First"
url: "https://bizztor.com/news/claude-duet-anthropic-proactive-ai-agent/"
author: "Bizztor"
author_url: "https://bizztor.com/author/bizztor/"
date_published: "2026-10-21T20:40:34.625Z"
date_modified: "2026-10-21T20:40:34.706Z"
category: "News"
description: "Claude Duet runs on Opus 5.5, checks your apps every few hours and asks before it pays or sends. It is US-only for now, so UK founders will have to wait."
publisher: "Bizztor"
---

# Anthropic Launches Claude Duet, an AI Agent That Messages You First

> Claude Duet runs on Opus 5.5, checks your apps every few hours and asks before it pays or sends. It is US-only for now, so UK founders will have to wait.

**Anthropic has launched Claude Duet, a separate Claude app that messages you first instead of waiting for a prompt.** The [research preview](https://claude.com/resources/articles/claude-duet) opened on 21 October 2026 through a waitlist, and only for Claude Pro and Max subscribers in the US.

Duet runs on Claude Opus 5.5 and can reach more than 6,000 connectors and plugins. It keeps working on its own cloud computer after you close the app, and it asks before it sends, pays, signs or deletes anything.

There is no extra charge. The catch is that Duet's background work comes out of the same usage allowance as the rest of your Claude plan, and that one decision shapes how the whole product behaves.

## What is Claude Duet?

Claude Duet is a proactive AI agent from Anthropic, the San Francisco company behind the Claude models. In Anthropic's launch announcement, the pitch is an assistant that decides what needs you, what can wait and when to check in before acting.

It is aimed at professionals and small business owners with more on their plate than they can track. Duet is a standalone app on iOS and Android, and on a computer you open it through claude.ai.

You don't start from scratch. Duet pulls in the memory, skills and connected apps already on your Claude account, so the first message arrives with context.

## How does Claude Duet work day to day?

You talk to Duet in one ongoing thread, typed or by voice, so a quick question and a week-long project start in the same place. [9to5Mac reports](https://9to5mac.com/2026/10/21/anthropic-takes-on-chatgpt-dots-and-meta-muse-with-claude-duet/) you can also open separate channels for different topics, and they share one memory.

Each morning Duet sends a short brief on what matters across your connected apps. Through the day it checks those apps and only messages you when something needs a decision.

Bigger jobs run in the background on a virtual machine with its own browser, so your laptop and phone can be switched off. Anthropic's examples run from booking meetings to building dashboards, sales forecasts, prototypes and web apps.

A Tasks tab tracks ongoing work, and finished files such as a deck or financial model land in a Make tab. 9to5Mac adds that, under Claude's support documentation, only you can mark a task as done.

## Why Duet checks your apps every few hours, not nonstop

This is the detail most launch coverage skips. Duet checks connected accounts every few hours, because Anthropic says checking constantly would burn through a plan's limits much faster.

Every background check counts against the allowance you also use for chat, coding and everything else in Claude. Larger tasks, like building a financial model, need more model turns and more background work, so they cost more of your plan.

The trade-off is written on Anthropic's own product page: Duet can miss something that arrives between check-ins. You can tell it what you want to hear about and how often, but more frequent checks will eat more usage.

OpenAI went the other way with Dots, its always-on agent launched at DevDay on 29 September for ChatGPT Pro and Business Premium users. [SiliconANGLE reported](https://siliconangle.com/2026/09/29/openai-launches-dots-always-on-ai-agents-in-chatgpt-with-their-own-cloud-computers/) that OpenAI's help documentation says agent usage won't count against plan allowances for the first month.

## What Duet will and won't do without you

Duet reads, drafts and prepares work on its own. Before it sends a message, makes a payment, signs a document or deletes anything, it shows you exactly what it plans to do and waits for your approval.

Some actions are blocked even if you say yes. Anthropic names trading securities and moving money between accounts. You can also disconnect any app Duet has access to at any time.

Purchases run through Link, Stripe's digital wallet. Duet confirms the item, Link asks you to approve the spend, and the payment goes on a virtual card so your real card number is never shared. It can also check out with Shopify's Shop Pay, and signing in to sites with logins saved in 1Password is listed as coming soon.

Agent payments are moving fast, though trust is lagging behind. Bizztor recently looked at [why most UK shoppers won't let AI agents buy for them](https://bizztor.com/insights/agentic-commerce-ai-agents-uk-shoppers/), and the same hesitation will apply to business spending.

## How Anthropic says it keeps Duet safe

Duet launches with a new classifier and safety system prompt that check each message for leaks. Requests to share card numbers or government ID are refused outright. Health or personal finance details go out only when you have explicitly told Duet to send them.

Its handling of credentials follows the Claude Managed Agents design, so Duet never directly sees your passwords or sign-in tokens. Actions pass through the auto mode classifier already used in Claude Code, and its defences against prompt injection come from Anthropic's browser-use research.

Red teams and an external firm penetration tested the app, and it is covered by Anthropic's public bug bounty. The whole approach draws on Anthropic's [framework for trustworthy agents](https://www.anthropic.com/research/trustworthy-agents), which also guides Claude Code and Claude in Chrome. Anthropic says it will publish more on Duet's safety design in the coming days.

## Which apps does Claude Duet connect to?

Duet uses the connectors and plugins in the Claude Marketplace. Anthropic's examples include finding missed Gmail and Slack messages, checking Intuit QuickBooks or Stripe for unpaid invoices and drafting reminders, and preparing payroll in Gusto with hours and bonuses filled in.

It can also draft a job post and publish it to LinkedIn once you approve, send agreements for signature through Docusign and write follow-ups from HubSpot deal history. Canva campaigns and Shopify store management are on the list too, with a richer Shopify visual editor promised soon.

Invoice chasing is a real pain point for small firms, and it is already a funded category in the UK, as [Adfin's $18M Series A for SME finance](https://bizztor.com/news/adfin-raises-18m-series-a-to-build-agentic-finance-platform-for-uk-smes/) shows.

## How early users are putting Duet to work

Anthropic's own staff tested it first. Boris Cherny, who leads the Claude Code team, talks through his mental to-do list with Duet on walks. Chelsea Larsson, who leads model and product behaviour design, hands it her list before time with her kids and gets an evening report.

Karan Ahuja, who leads the Duet team, uses it to schedule user interviews, build prototypes and even ship code for Duet itself.

Small business testers give a more practical picture. Jess Provencher, broker and owner at Pro Homes, had Duet flag a property showing that sat on her calendar for the wrong day, without being asked.

Amin Ebrahimi owns Shahs of Kabob in South Florida, which has about 110 employees. He has used Duet for accounting, customer support and marketing, and worked through his invoices store by store. "I don't need to open up 100 tabs," he told Anthropic.

Brand strategist Mya Dalton had Duet prepare a brief before a call with a potential client. It also noticed the call fell right before her flight and offered to move it.

## What isn't ready yet

Anthropic is upfront that this is an early build. Duet may misread a thread or get urgency wrong, and personal errands such as booking tickets are weaker than work tasks. You also can't watch its browser yet.

For now it lives apart from the main Claude app. Anthropic says feedback from early users will decide how the two eventually fit together.

## How Claude Duet compares with ChatGPT Dots and Meta Muse

9to5Mac framed the launch as Anthropic's answer to OpenAI's Dots and Meta's Muse, a personal agent Meta released in the US. All three chase the same idea: an assistant that carries on when you stop typing.

The design choices differ. Dots live inside ChatGPT and can be reached from Slack and Teams. Duet sits in its own app, inherits your Claude setup and leans hard on approvals for anything that leaves your control.

## What Claude Duet means for UK founders

UK founders can't use Duet yet. Anthropic has given no date for other countries or for Team and Enterprise plans, and 9to5Mac reports invitations will go out in waves.

The tools in Anthropic's demos also lean American. Gusto, for one, is a US payroll service, so Bizztor would want to see how Duet handles UK payroll and accounting before trusting it with the books.

The useful move now costs nothing. Duet inherits your Claude memory, skills and connectors, so founders already on Pro or Max who tidy those up today will get more from it on day one. Bizztor's guide to [AI tools that help small teams and solo founders do more](https://bizztor.com/guides/5-ai-tools-that-let-small-teams-and-solo-founders-do-more/) is a sensible starting point.

The real test is judgement. An agent that messages you first is only worth having if it flags the right things, and with [Anthropic's agents already logging six hours for every human hour](https://bizztor.com/news/anthropic-ai-agents-outpace-human-hours-junior-hires-drop/), that question is coming for UK businesses whether Duet arrives this year or next.

US Pro and Max subscribers can join the waitlist on the Claude Duet product page today.

---

Source: https://bizztor.com/news/claude-duet-anthropic-proactive-ai-agent/
