Published on October 22, 2026 at 10:26 AM

Anthropic unveils Claude Duet, an AI agent that works for you like Muse (Meta) and Dots (OpenAI)

AI at your service

3 min

Nicolas Lellouche

{"@context":"https:\/\/schema.org","@type":"Person","@id":"https:\/\/www.numerama.com\/#\/schema\/person\/e1c5c505bb96ef4588035c09ccaf93e2","name":"Nicolas Lellouche","url":"https:\/\/www.numerama.com\/author\/nicolaslellouche\/","description":"Fascinated by new technologies since childhood, Nicolas first joined the editorial team of 01net in 2017 before becoming head of Numerama's tech section in 2022. On his desk are countless smartphones, lengthy photocopies of European texts, and AI agents awaiting instructions. You can also see him on the program Télématin on France 2.","sameAs":["https:\/\/x.com\/LelloucheNico","https:\/\/www.instagram.com\/nico_lellouche\/","https:\/\/fr.linkedin.com\/in\/lellouchenico"],"image":{"@type":"ImageObject","url":"https:\/\/img.numerama.com\/www.numerama.com\/wp-content\/uploads\/2026\/07\/nico-200x200.jpg?resize=96,96\u0026key=9f405155"}}

3 min

3 min

Nicolas Lellouche

No longer see this ad

Zen Reading

Summarize the article

AI-generated summary, verified by Numerama

We generate a summary of the article for you using AI, and we verify it to make sure there are no errors, in order to guarantee the authenticity of the information.

Learn more

Anthropic launches Claude Duet, an AI personal agent application that resides in the cloud, capable of managing tasks and summarizing key information daily.
Unlike traditional chatbots, Claude Duet can initiate contact and integrates with various services, supporting professional activities such as email management and meeting scheduling.
Currently available in the United States only, Claude Duet runs on a separate virtual machine, reflecting a cautious approach to privacy concerns and restrictions on access to personal data.

Receive a summary of the key news every evening with Le Récap’

window._rmxd={theme:document.documentElement.getAttribute('data-scheme')||'light'};

Anthropic launches Claude Duet, a new app where Claude becomes a personal assistant. This AI agent monitors your apps, sends you a summary every morning, and runs tasks in the background from a virtual machine in the cloud.

After Meta and OpenAI, Anthropic is entering the personal AI agents space. The creator of Claude is not entirely unfamiliar with agentic features: Claude Cowork and Claude Code already enable the automation of many tasks and have often been regarded as the industry benchmarks. But Anthropic now wants to go further with a dedicated app: Claude Duet.

In a blog post on October 21, Anthropic announced Claude Duet, a new app with an agent capable of working without human prompting. For now, it is a research preview available by waitlist: Anthropic is moving more slowly than its competitors and emphasizes safety.

We’re opening a research preview of Claude Duet, a new app where Claude works as a proactive agent.

It comes to you with what needs your attention, completes tasks you ask of it, keeps working when you step away, and checks with you before it acts on your behalf. pic.twitter.com/kt8gAbWuX8
— Claude (@claudeai) October 21, 2026

After Muse and Dots, here comes Claude Duet

Anthropic makes a surprising choice: for the time being, Claude Duet is not integrated into the Claude app. The service takes the form of a separate app, available on iOS and Android, as well as on the web via claude.ai. However, it does integrate directly with Claude accounts: Duet retrieves the chatbot’s memory, skills, and connected apps.

The difference from a traditional chatbot is that Duet can take the initiative to contact its users. Anthropic is targeting professionals and small business owners and acknowledges that personal tasks, such as booking tickets, are currently handled less effectively. The examples given are:

Send a summary of the important items in your connected apps every morning, then message you during the day only when something requires your attention.
Catch up on the Gmail and Slack messages you missed and draft replies.
Identify unpaid invoices in QuickBooks or Stripe and draft reminders.
Schedule meetings, prepare payroll, post a job listing on LinkedIn, or send a contract for signature via Docusign.
Build dashboards, financial forecasts, prototypes, or web apps.
Buy products for your business.

Claude Duet prend la forme d'une conversation.

Claude Duet takes the form of a conversation. // Source: Anthropic

Another difference from a traditional chatbot: Duet operates entirely in the cloud. Anthropic gives it a virtual machine with its own browser, enabling it to continue working when your computer and phone are turned off. Duet never has direct access to the computer: Anthropic wants to avoid the criticism received by its competitors.

Anthropic is coming after its competitors.Meta launched Muse on September 8 in the United States: an agent that lives in its own virtual machine and is accessible from an app, the web, or WhatsApp, with a free offering and subscriptions for $20 and $100 per month. OpenAI responded on September 29 with the Dots, agents “always active” powered by GPT-6 Astra, reserved for ChatGPT Pro and Business Premium subscribers, and offered in beta to Enterprise plans. All three rely on the same idea: a proactive agent, with its own computer in the cloud, that works on your tasks while you are away.

A cautious launch, amid mistrust

Like Muse at launch, Claude Duet is restricted to the United States. You must be subscribed to Claude Pro or Max and sign up for a waiting list: Anthropic will send an email when access is ready. Duet is not billed separately, but it uses your subscription’s usage limits, including for its background checks. This is also why it only checks your accounts every few hours. No date has been announced for France.

Not everyone is convinced that these agents represent the future. Privacy specialists point out that an agent is useful only if it accesses your data, and that you cannot erase what it has already seen by revoking its access. Apple and Microsoft could gradually strengthen controls to prevent AIs from siphoning users’ personal data.