Anthropic opened a waitlist for Claude Duet on October 21, a background agent that watches a user's connected work apps, sends a morning brief and does tasks on its own, but stops to ask before it sends a message, pays, signs a document or deletes anything. Anthropic's announcement calls it a research preview for US subscribers to its Pro, Max and Ultra plans, built around Claude Opus 5.5.
Access is through the waitlist: Anthropic says it will email people on the waitlist when their turn comes and gives no timeline. Duet arrives 22 days after OpenAI's dots and 43 days after Meta's Muse, two agents that also run on their own cloud computers and keep working after the user closes the app.
What Duet does
Duet runs on a virtual machine with its own browser, so a user's laptop or phone does not need to stay on. It checks connected accounts every few hours, flags what needs attention and takes on work from scheduling meetings to building dashboards, financial or sales forecasts, prototypes and web apps, according to the announcement. Users talk to it by text or voice in one running thread, in a standalone iOS and Android app or at claude.ai.
It carries over the memory, skills and connected apps a user already has in Claude. Anthropic says it can draw on more than 6,000 connectors and plugins from the Claude Marketplace and names Gmail, Slack, QuickBooks, Stripe, Gusto, LinkedIn, DocuSign, HubSpot, Canva and Shopify. Saved 1Password logins and a visual storefront editor for Shopify are listed as coming soon.
Anthropic says it built Duet for professionals and small business owners. Team and Enterprise plans are not included yet, according to the product page.
Where it stops and asks
Before Duet sends a potentially sensitive message, pays for something or deletes anything, it shows the action, the destination and the content, including every recipient, and waits, according to a safety and privacy note Anthropic published on October 23. Other limits from that note:
Trading securities, moving money between accounts and using someone else's account are blocked even if the user approves.
Paying, sending money, deleting an order and creating an account can never be set to "don't ask me again".
Purchases go through a one-time card from a wallet such as Link by Stripe, limited to one merchant and one amount, so the real card number is not shared.
Duet never sees passwords, cookies or sign-in tokens, and a separate classifier blocks credentials, full card and bank numbers and government ID numbers from outgoing messages.
Conversation history and memory stay until the user deletes or resets them. Disconnecting an app stops new access but does not erase what Duet already learned.
Anthropic is also plain about what does not work yet. Personal errands such as booking tickets are weaker because the product was built around work, and Duet cannot yet show the user its browser.
It still might misread a thread or miscalibrate what's urgent.Anthropic, in its Claude Duet announcement
Duet, dots and Muse side by side
All three agents, listed with their launch dates below, run on their own cloud computer and check in before sensitive actions. The differences are in who can get them and what they cost.
Agent | Plans | Where |
|---|---|---|
Duet, Oct. 21 | Pro, Max, Ultra | US, waitlist |
dots, Sep. 29 | Pro, Business Premium, Enterprise | Pro: not EEA, UK, Swiss |
Muse, Sep. 8 | Free, paid tiers | US |
Sources: Anthropic, OpenAI's dots documentation, Meta. The dots launch date as reported by TechCrunch.
What it means for buyers
For an individual, Duet costs nothing beyond a Claude subscription: it comes with Claude Pro, which costs $20 a month, or $200 a year paid up front. Max is $100 a month and Ultra $200, five and 10 times the Pro price. Meta's Muse is free for most use, with paid plans for more, and Meta does not list a price on its launch post.
The catch is usage. During the preview Duet shares the plan's limits with the rest of Claude, background checks included, and Anthropic says larger tasks such as financial models or decks take more of them. OpenAI's documentation says conversations with a dot do not count toward ChatGPT limits. A Pro subscriber who already runs into Claude's limits will find Duet competing for the same allowance.
For companies, the question is control. Duet is sold only on individual plans for now, so a small business owner can sign up, but an IT team cannot turn it on or off for staff the way OpenAI lets Enterprise admins do with dots, which ship switched off by default. Until Team and Enterprise are added, any Duet in a workplace runs on an employee's personal subscription, with that person's own app connections and memory.
The safety design is close to Meta's: both check out through Link by Stripe and both list 1Password logins as coming. Where Anthropic goes further on paper is the published list of actions Duet refuses outright and the steps a user can never pre-approve. Anthropic says how often Duet asks for approval is still changing, so a buyer testing it now is testing a preview, which is what Anthropic calls it.
For earlier coverage, see SaasBuzz on Anthropic's Ultra plan and the pushback on Meta's Muse.
