Claude Duet is Anthropic’s attempt at an assistant that does not wait to be asked. On Wednesday the company opened a research preview of the standalone app, in which Claude works as a proactive agent: it messages the user when something needs attention, carries out tasks in the background and checks before acting on the user’s behalf, according to Anthropic’s announcement. Access is by waitlist, starting with Pro and Max subscribers in the United States, on iOS, Android and the web.
What Duet does
An AI agent is software that can take steps on its own, such as opening apps, reading messages or filling in forms, rather than only answering questions. Duet sends a short brief each morning on the most important items across a user’s connected apps. During the day it checks those accounts every few hours and writes only when something needs a response. Anthropic says it does not check constantly because that would use up a plan’s limits much faster.
Users can also hand it work, from scheduling meetings to building dashboards, sales forecasts and simple web apps. Duet runs on Claude Opus 5.5 and has its own virtual machine with a browser, so tasks keep going when the user’s phone or laptop is off. Everything happens in one ongoing thread, by text or voice, and Duet sends a notification when a task is done or needs review.
Setup is light because Duet carries over memory, skills and connected apps from the user’s existing Claude account. Anthropic says it can use more than 6,000 connectors and plugins, including Gmail, Slack, QuickBooks, Stripe, Gusto, Docusign and HubSpot. It can also buy things for a business through Stripe’s Link wallet, paying with a virtual card so the real card number is not shared.
Where Anthropic draws the line
An agent that reads email and sends messages is a security question as much as a convenience. Anthropic says Duet reads, drafts and prepares work on its own, but before it sends a message, makes a payment, signs a document or deletes anything, it shows the exact action and waits for approval. Trading securities and moving money between accounts are off-limits even with approval.
The company says Duet never directly sees passwords or sign-in tokens. A new classifier, a separate model that screens each outgoing message, is designed to block leaks of the most sensitive data, including credit card numbers and government ID. Health and personal finance details go out only when the user has explicitly told Duet to send them. Red teams and an outside firm penetration-tested the product, and it is covered by Anthropic’s public bug bounty program. Anthropic says it will share more detail on the safety design in the coming days.
Who it is for
Anthropic is aiming Duet at professionals and small business owners. Its examples are practical rather than dramatic. A real estate broker said Duet flagged a property showing that was on her calendar for the wrong day. A brand strategist said it noticed a client call scheduled right before a flight and offered to move it. The owner of a restaurant chain with about 110 employees said he has used it for accounting, customer support and marketing.
For a small business, the pitch is that one assistant can watch invoices in QuickBooks, payroll in Gusto and leads in HubSpot without anyone opening each app. The value depends on how well it judges what matters, which Anthropic itself calls the core of the product.
A crowded field
Duet arrives three weeks after OpenAI introduced Dots, always-on agents that run on their own cloud computer and can be messaged or called, at its developer event on September 29, the San Francisco Standard reported. Meta launched its own agent, Muse, in September, and it climbed to second place on the U.S. App Store. Media coverage has framed Duet as Anthropic’s answer to both, iPhone in Canada noted.
What is still unproven
Anthropic is open about the limits. Duet may misread a thread or misjudge what is urgent. Personal errands such as booking tickets are weaker than work tasks, and the app cannot yet show users its browser. It lives apart from the main Claude app for now.
Usage is the other open question. Background checks count against plan limits, and large jobs such as financial models or slide decks take more model turns. How quickly Duet uses up a Pro subscription will matter as much as how capable it is.
Our take
The interesting part of Duet is not that it can do tasks, since several agents can, but that it decides when to speak up. That is a harder problem: an agent that interrupts too often becomes noise, and one that stays quiet is just a chatbot. Anthropic’s approval rules for payments, signatures and deletions are the right default for a product that touches a business’s accounts. Small business owners should try it on low-stakes work first and watch how much of their plan it uses before trusting it with anything that costs money.