
Anthropic opens a preview of Claude Duet, an assistant app that comes to you first
Claude Duet, an assistant that speaks up first
- Flags what needs you
- Does what you ask
- Keeps working while you're away
- Asks before acting for you
Anthropic opened a research preview of Claude Duet around 4 AM KST on October 22. In this app Claude doesn't wait to be asked. It brings you what needs your attention, finishes the tasks you hand it, keeps going when you step away, and checks with you before it acts on your behalf. It runs as an app on iOS and Android and on claude.ai.
Source: Anthropic announcement ↗
US Pro and Max subscribers first, by waitlist
For now, only Pro and Max subscribers in the US can join the waitlist. Anthropic admits that this early version can still misjudge what is urgent. It plans to keep tuning Duet based on what early users tell it.
Source: Claude thread on X ↗
Duet shares your existing Claude usage limits
During the preview, Claude Duet is included in the Pro and Max plans and shares the same usage limits as the rest of Claude. It uses them differently from a normal chat, though. Its background checks on your connected accounts count toward the limits, and big jobs such as a financial model or a deck take more model turns. Anthropic made it check every few hours, because checking nonstop would drain the limits much faster.
Source: Anthropic announcement ↗
Approval before sending, paying, signing or deleting
Shows the action, then waits for approval
- Send a message
- Pay
- Sign a document
- Delete
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 is about to do and waits for your approval. Some actions stay off limits even with approval, such as trading securities or moving money between accounts. It never directly sees your passwords or sign-in tokens.
Source: Anthropic announcement ↗
Anthropic opened a research preview of Claude Duet early on October 22 (KST). Duet checks your connected apps every few hours and messages you first when something needs you. It carries on with tasks in its own virtual machine, and it asks for approval before anything hard to undo, such as sending or paying. US Pro and Max subscribers get it first through a waitlist, and it draws on the same usage limits as the rest of Claude.
What shipped
- It launched at 04:09 KST on October 22, 2026 (19:09 UTC on October 21). The official Claude account posted at that time, and the Anthropic article is dated October 21. (Claude on X, Anthropic announcement)
- It is built for professionals and small business owners with too much to track. Anthropic pitches it as a partner that helps you decide where to spend your time and energy.
- It sends a short brief every morning. The brief covers the most important items across your connected apps. During the day it watches those apps and messages you only when something needs your attention. You can tell it what you want to hear about and how often.
- It is designed around Claude Opus 5.5. Anthropic says the model choice is about judgment: deciding what needs you, what can wait and when to check with you first. TestingCatalog reported the same right after the launch. (TestingCatalog)
- It brings over the memory, skills and connected apps from your Claude account. Setup takes very little work. You talk to it by typing or by voice in one ongoing thread, so a quick reply and a weeklong project start in the same place.
- It works in its own virtual machine with its own browser. Your laptop or phone does not need to stay on. Describe a task in a short message, and it works out which apps to use and which steps to take, then notifies you when the task is done or needs review.
- It can use the connectors and plugins in Claude Marketplace. Examples in the announcement include replying to missed Gmail and Slack messages, finding unpaid invoices in QuickBooks or Stripe and drafting reminders, preparing payroll in Gusto, sending agreements for signature through Docusign and managing a Shopify store.
- It can buy things for your business. It pays with Link, Stripe's digital wallet. Duet confirms the item with you, Link asks you to approve the spend, and a virtual card pays so your real card number is not shared. It can also check out with Shopify's Shop Pay. Signing in to sites with logins saved in 1Password, with your permission, is coming soon.
- For now it lives in its own app, separate from the main Claude app. Anthropic says early feedback will shape how the two work together.
How it handles sensitive information
- It never directly sees your passwords or sign-in tokens. Credentials follow the same design as Claude Managed Agents.
- A classifier checks each action. Duet uses the auto mode classifier that stops Claude and Claude Code from taking steps you didn't ask for.
- A new classifier guards against leaks. A dedicated classifier and a safety system prompt check every message. Duet refuses to share the most sensitive tier of information, such as credit card numbers and government IDs. Health or personal finance details go out only when you explicitly tell it to send them.
- Its defenses against hidden instructions and prompt injection come from Anthropic's browser use research. Red teams and an outside firm ran penetration tests, and Duet is covered by the public bug bounty program.
- Important decisions stay with you. You can disconnect any app at any time. Anthropic says it will share more about Duet's safety design in the coming days.
Launch-day integrations
- Canva. Turns your tasks, meeting notes and more into social posts, designs and presentations that you can edit in Canva. Duet asks before it sends anything. (Canva)
- Airtable. Pulls updates from email, Slack and shared documents to keep your records current. You can check and update your base from your phone. (Airtable)
Reactions
Lydia Hallie said Duet has made focused work much easier. She has it watch Slack, email and her calendar and ping her only when something matters, so she no longer has to be "on" all the time and switches context far less. It even covers threads she isn't in, and it often attaches a doc or design for context.
Claude Duet has made focused work so much easier! Not having to be "on" all the time is so nice, way less context switching My setup is boring lol but I have it watch Slack/email/cal and only ping me when it matters (even threads I'm not in), often with a doc/design for context!
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.
Many posts on X compared it with Grok Bot, Meta Muse and ChatGPT Dots. (example)
Still unclear
- There is no date yet for a launch outside the US.
- Anthropic has not said how much of the limits the background checks use. The article only says it checks every few hours.
- The details of the safety design are not out yet.
My take
I want to try it, so it's a shame it's US only. And since it shares the same limits, I'm worried about what it will cost.
I use Grok Bot every day. I wrote up how I split work across it in How a Solo Developer Uses Grok Bot. While using Grok Bot, I had wished for an agent like it built on Claude. Right now I pay for a Grok plan only because of Grok Bot. If Claude Duet gives me a similar experience, that alone could be enough reason to switch.
What I like most about Grok Bot is that I can create a separate bot for each job, and that I can hand coding work to Cursor's cloud agents. For Claude Duet to be a real reason to switch, it would need to support working like that too.
