What is Claude Duet? Anthropic’s proactive AI agent explained

Anthropic has opened a research preview of Claude Duet, a new app in which Claude works as a proactive agent: instead of waiting for you to type a prompt, it comes to you with what needs attention, works on tasks while you are away and asks before it acts for you. The company announced it on 21 October and says it built Duet for “professionals and small business owners with more on their plates than they can track”.

It is Anthropic’s answer to OpenAI’s Dots and Meta’s Muse, which arrived earlier this autumn. Here is what Duet does, who can get it and what Anthropic says about safety.

How Claude Duet works

• A morning brief. Each day Duet sends a short summary of the most important items across the apps you have connected.
• Messages only when needed. During the day it keeps an eye on those apps and pings you when something needs you, rather than constantly.
• Tasks of any size. You can hand it simple admin such as scheduling meetings, or bigger jobs such as building dashboards across many tools, drafting sales or financial forecasts, prototyping and building web apps.
• One ongoing thread. You talk to it by typing or by voice in a single conversation, so a quick reply and a week-long project start in the same place.

Anthropic says Duet is built around Claude Opus 5.5 and is designed to judge what needs you, what can wait and when to check first. It also carries over your existing memory, skills and connected apps from your Claude account, so setup is meant to be light.

It keeps working when you put the phone down

Duet runs on its own virtual machine with its own browser, so your laptop or phone does not need to stay on. You describe a task, it works out which apps and steps to use, and you get a notification when it is done or needs your review. To save your plan’s limits, Anthropic says Duet checks connected accounts every few hours rather than nonstop, and Engadget reports you can tune how often it messages you and about which topics.

Who can get it

For now, Duet is a waitlist-only research preview, starting with Pro and Max subscribers in the US. It is available as standalone apps for iOS and Android and on claude.ai, and it uses the same usage limits as the rest of Claude, including Claude Code. Engadget notes that this sets it apart from OpenAI’s Dots, which it says is limited to OpenAI plans starting at $100 a month; Claude Pro costs $20 a month.

What Anthropic says about safety and privacy

An assistant that reads your email and sends messages for you is constantly deciding where information goes, and Anthropic lists several safeguards:

• No direct access to credentials: Duet never sees your passwords or sign-in tokens, following the design of Claude Managed Agents.
• Action checks: an automatic classifier approves actions to stop steps you did not ask for, and Duet asks before acting on your behalf.
• Prompt-injection defences drawn from Anthropic’s browser-agent research.
• A new sensitive-data classifier that refuses to share the strictest information, such as card numbers and government ID, and only passes on health or financial details when you explicitly tell it to.
• External testing: red teams and an outside firm ran penetration tests, and Duet is covered by the bug bounty programme.

Anthropic says it will publish more detail on Duet’s safety design in the coming days. The usual caution applies: the more accounts you connect, the more an agent can do, and the more a mistake can cost. Our coverage of OpenAI agents that switched off their own safety monitoring and the Meta Muse share-link exposure shows why that matters. If you would rather keep AI on your own hardware, see our laptop guide for local AI.