Anthropic Launches Claude Duet, an AI That Works While You're Away
It keeps an eye on your connected apps and keeps working while you are away. A research preview has opened with a waitlist, starting with Pro and Max subscribers in the US.

Anthropic unveiled Claude Duet, an AI app for work, on October 21 local time. It is a research preview ahead of a general release. The app checks the apps you have connected on your behalf, flags what you need to know, and handles tasks you hand off to it. Pro and Max subscribers in the United States can join the waitlist first.
It tells you before you ask
Claude Duet is designed to stay on top of your to-dos before you think to ask. Anthropic says it is built on Claude Opus 5.5 and tuned to sort what needs your attention now from what can wait. The main audience is working professionals and small business owners juggling several tasks at once.
Every morning it sends a short digest of the key items from your connected apps. After that, it checks your accounts every few hours and alerts you when something comes up that you should know about. You decide what kinds of updates you get and how often. According to the company, you can also hand off scheduling meetings, revenue forecasting, and even building web apps.
It carries over the memory built up in your existing Claude account, your Skills (a feature that stores how you like work done), and the apps you have already connected. You give instructions in a single chat window, by text or voice. The work itself runs on a virtual computer with its own browser, so tasks keep going even when your phone or laptop is off. It sends a notification when a job is finished or when it needs your review.
Payments and outgoing messages require your approval
Connecting to your work tools is the core of the product. Anthropic says you can use more than 6,000 connectors and plugins available in the Claude Marketplace. It can draft replies to Gmail and Slack messages you missed, or track down unpaid invoices and draft payment reminders. Writing and posting a job listing, and preparing a contract to send out for review and signature, were also cited as use cases.
Any action that affects the outside world requires your approval. Before sending a message, making a payment, signing a document, or deleting something, it shows you what it plans to do and waits. Securities trading and moving money between accounts are off limits even if you approve them. When buying business supplies through Link, it confirms the items and gets spending approval before paying with a virtual card.
The company says it has added separate safeguards around sensitive information. It refuses requests to share credit card numbers and government-issued ID details. Health information and personal financial information are only passed along when you explicitly instruct it to send them. That said, Anthropic notes that this early version can misread the intent of a conversation or misjudge how urgent something is.
Every use eats into your plan limits
The research preview is included with the Pro and Max plans. It is available as a separate iOS and Android app, installed apart from the existing Claude app, and on the web at claude.ai. Usage limits are shared with other Claude features. Periodic app checks count against your limits, and larger jobs like building financial models or presentation decks consume more.
Among the early examples Anthropic shared: Duet caught a real estate showing that had been scheduled incorrectly. In another, it suggested moving a call with a prospective client that had been booked right before a flight departure. The open questions are how accurate those judgment calls turn out to be, and how useful repeated checks and task handling will be within the plan limits.