Claude Duet: What Anthropic's Proactive Agent Actually Does, and Who Can Try It

Quick answer: Claude Duet is Anthropic's proactive agent app. It sends a morning brief, works on tasks in the background on a cloud computer, and asks for approval before it sends a message, pays, signs a document or deletes anything. It opened as a research preview on October 21, 2026 for Pro and Max subscribers in the US, by waitlist, at no extra cost.

Two details decide whether Duet matters to you this week. The first is geography. If you are not on a paid Claude plan in the United States, you cannot get in, and Anthropic has not said when that changes. The second is the shape of the product. Duet is a separate app built on the assumption that Claude should start the conversation. That changes what you type, what you check, and what can keep running after you close the laptop.

I have not used Duet. This post is a read of Anthropic's launch material, the product page and the FAQ, with screenshots taken from those pages on October 21, 2026. Where a claim comes from a third party, I say so.

Claude Duet at a glance

Question	Answer as of October 21, 2026
What it is	A standalone app where Claude works as a proactive agent
Maker	Anthropic
Status	Research preview, opened October 21, 2026
Model	Claude Opus 5.5
Platforms	iOS, Android and claude.ai
Price	Included in Pro and Max during the preview, no separate fee
Access	Waitlist, Pro and Max subscribers in the US
Control	Approval before sending, paying, signing or deleting

What Claude Duet is

Claude Duet is a personal AI agent you message. It reads the apps you connect, decides what needs you, and reaches out on its own. Anthropic's own comparison is short: Claude waits, Duet comes to you. You can hand it work by text or voice in one ongoing thread, and it runs that work on a virtual machine with its own browser.

That last part is the difference between an agent and a chat window with a to-do list. Duet does not need your laptop open, and it does not need you watching. You describe the task in a message, it works out which apps to use, and it notifies you when the job is done or needs your review.

Anthropic says it built the product for professionals and small business owners with more on their plates than they can track. The examples lean that way: invoices, payroll, contracts, hiring, a storefront, a launch campaign. So does the waitlist form, which asks what you would hand off first.

The morning brief is the whole pitch

Each morning, Duet sends a short brief on the most important items across your connected apps. Through the day it keeps tabs on those apps and messages you only when something needs attention.

The demo above is Anthropic's illustration, not a customer's real inbox, and it is worth reading closely because it shows what the product is trying to be. The items are not tasks someone typed into a project manager. They are things the agent noticed: a contract that came back redlined, a permit that needs a signature and $640, a customer issue that already has refunds drafted. The brief ranks them and says what has been prepared.

That is a narrower promise than "AI runs your business", and a more useful one. Jess Provencher, a broker at Pro Homes, told Anthropic that Duet once flagged a property showing that was on her calendar for the wrong day, without being asked. Mya Dalton, a brand strategist, said Duet noticed a client call fell right before her flight and offered to move it. Neither catch was a task anyone assigned.

How the background work and approvals fit together

Duet runs on a cloud virtual machine with its own browser. You send a short message, it works out the steps, and it comes back when the task is finished or needs a decision. Anthropic says approvals, logins and safety checks run in that same cloud computer, so the sensitive parts stay out of the chat.

The product page splits the design into two promises, and the screenshot below shows both columns.

On the left, the claims are about judgment and continuity: the brief, the large tasks, the memory and skills you already have in Claude. On the right, they are about control: you approve sensitive operations, safety checks run in your cloud computer, and you can see what Duet has done and is doing.

Before Duet sends a message, makes a payment, signs a document or deletes anything, it shows you what it is about to do and waits. Two things are off-limits even if you say yes: trading securities and moving money between accounts. You can also disconnect any app at any time.

An approval prompt is only as good as the person reading it. If you approve a payment without checking the amount, the safeguard did its job and you did not. Anthropic calls this an early version, and it is right to.

What Claude Duet costs

There is no Duet subscription. During the research preview it is included in Claude Pro and Max plans and shares their usage limits.

The plans, per Anthropic's pricing page on October 21, 2026:

Plan	Price	What Duet gets from it
Pro	$20 a month, or $17 a month billed annually at $200 up front	Duet access during the preview, sharing the plan's limits
Max	From $100 a month, 5x or 20x Pro usage	The same, with a larger allowance to draw from

The usage part is the bit people skip. Duet's background checks of your connected accounts count toward your limits, and bigger tasks take more model turns than a chat reply. Anthropic says it checks every few hours rather than nonstop specifically because constant checking would burn through limits faster. If you already hit your Pro limits most weeks, expect to hit them sooner with Duet running. We went through the plan math in our Claude Code pricing breakdown, and the same logic applies here.

Who can use it, and who cannot

US Pro and Max subscribers, by waitlist. Team and Enterprise plans are not included yet. Anthropic's announcement does not commit to a date for either, and it does not name a second country.

The waitlist form asks for a first name, last name, email, company name, phone type (iPhone or Android) and what you would hand off to Duet first. Anthropic says it will email when Duet is ready for your account.

If you are outside the US, the practical answer is to join the waitlist only if the form accepts you, and otherwise wait for an announcement. There is no VPN story worth building a workflow on, and no published expansion date to plan around.

How Duet handles passwords, money and prompt injection

Anthropic's safety claims for Duet rest on work it has already published for other agents.

• Duet follows Anthropic's framework for trustworthy agents, the same one behind Claude Code and Claude in Chrome.
• Its credential handling follows Claude Managed Agents, so it never directly sees your passwords or sign-in tokens.
• Its actions are screened by the auto mode classifier that also protects Claude and Claude Code from steps you did not ask for.
• Prompt-injection defenses come from Anthropic's browser-use research, plus model-level safeguards.
• Red teams and an outside firm penetration tested it, and it is covered by Anthropic's public bug bounty.

New for Duet is a classifier and safety system prompt aimed at leaks. Anthropic says it rejects requests to share the strictest tier of information, including credit card numbers and government identification. Softer categories, such as health or personal finances, go out only when you have explicitly told Duet to send them.

Payments get their own layer. Duet can buy things through Stripe's Link: Duet confirms the item with you, Link asks you to approve the spend, and the payment runs on a virtual card so your real card number is not shared. Shopify's Shop Pay works the same way at checkout. A richer visual editor for Shopify storefronts and 1Password sign-in are marked coming soon.

Read all of that as Anthropic's description of its own design. None of it is an independent security audit, and an approval-first agent still depends on you reading the prompt.

What early testers did with it

Anthropic's announcement names three people inside the company and three small business owners. The internal examples are about attention: Boris Cherny, who leads the Claude Code team, talks through his to-do list with Duet on walks instead of saving it for his desk. Chelsea Larsson hands over her list before leaving to be with her kids and reads the results as an evening report. Karan Ahuja, who leads the Duet team, uses it to schedule user interviews, build prototypes and ship code.

The outside examples are more useful if you run a small company.

Amin Ebrahimi owns Shahs of Kabob in South Florida with about 110 employees. He used Duet for accounting, customer support and marketing, and worked through invoices store by store. His line to Anthropic was that it brings everything together and he does not need 100 tabs open. That is a real workflow for an owner with 110 staff and no operations team.

The pattern across all six stories is small catches and prepared work, not autonomy: a wrong date, a flight conflict, an invoice reminder, a brief before a call. That is what the product is selling, at least today.

Claude Duet vs OpenAI Dots, Meta Muse and Grok Bot

Four always-on agents arrived within about ten weeks of each other. They look similar in a demo and differ in the details that decide whether you can use one.

	Claude Duet	OpenAI Dots	Meta Muse	Grok Bot
Maker	Anthropic	OpenAI	Meta	xAI
Launch	Oct 21, 2026, research preview	Sept 29, 2026	Sept 8, 2026, US	Aug 11, 2026, beta
Where you use it	iOS, Android, claude.ai	ChatGPT, Slack, Teams	Meta's consumer app	macOS, iOS
Who gets it	US Pro and Max, waitlist	Eligible Pro, Business and some Enterprise plans	US consumers	SuperGrok and Cursor plans
Cost shape	Included in Pro and Max, shares limits	Tied to the ChatGPT plan	Consumer plans	Bot usage separate from Grok and Cursor
Control	Before send, pay, sign, delete	Sensitive actions need you	Purchases through Link by Stripe	Comes back when it needs approval

Sources for the other three: OpenAI's launch coverage, TechCrunch's Muse report and xAI's Grok Bot announcement, all linked at the bottom.

The differences that matter:

Dots vs Duet. Dots lives inside ChatGPT, reaches into Slack and Teams, and rolls out to eligible Pro, Business and some Enterprise plans, with the EEA, Switzerland and the UK excluded for personal Pro at launch, per launch coverage. Duet is a separate app, US-only, paid-plan only. If your team already lives in ChatGPT, Dots fits the habit better. If you already pay for Claude and use its memory and connectors, Duet arrives with your context.

Muse vs Duet. Muse is a consumer agent for email, calendar, payments, health and the smart home. Duet is aimed at work: invoices, payroll, contracts, hiring, a storefront. Anthropic says Duet is weaker at personal errands. They are chasing different buyers.

Grok Bot vs Duet. Grok Bot sells a team of named Bots, with a chief of staff on top and specialists underneath, and it reads like a product for teams that want several agents running at once. It also comes with its own usage allowance, separate from your Grok and Cursor plans, where Duet shares yours. Duet starts as one agent for one person. For the longer comparison, we covered Grok Bot and Meta Muse side by side.

Where Claude Duet falls short

Anthropic is unusually direct about the preview's limits, and they are worth taking at face value:

• It may misread a thread or miscalibrate what is urgent.
• It can miss something that arrives between check-ins, because it checks every few hours.
• It is built around work, so personal errands such as booking tickets are weaker.
• It cannot show you its browser yet, so you cannot watch it work step by step.
• It lives in its own app, separate from the main Claude app.
• It shares your usage limits, and background work draws them down.
• Access is US-only, Pro and Max only, by waitlist.

Two more limits are not on Anthropic's list but matter for planning. There is no API, so you cannot build Duet into your product. And there is no Team or Enterprise access, so a small company cannot roll it out to staff yet, even if the owner has a Max plan.

How to join the Claude Duet waitlist

1. Check that you have a Claude Pro or Max plan and a US account. Team and Enterprise plans are not part of the preview.
2. Open claude.com/product/claude-duet. The waitlist form is at the top of the page and again at the bottom.
3. Fill it in, including the question about what you would hand off first. Anthropic says it will email you when Duet is ready for your account.
4. Review the apps connected to your Claude account. Duet carries over your memory, skills and connectors, so you are handing it that context on day one.
5. When you get in, start with drafts, summaries and lookups. Widen access after you have watched a few approvals.

What I would hand it first

If you run a small SaaS, the useful test is a task that recurs, has a clear finish, and does not move money on the first pass. Three to try:

• The invoice chase. If your billing runs through Stripe or QuickBooks, ask Duet to find unpaid invoices and draft reminders for your review. The approval step is exactly where you want the human.
• Support triage. Connect the support inbox and ask for a morning summary of tickets that need a human, with draft replies for the easy ones. Read the drafts before they go anywhere.
• The weekly numbers. Ask for a Monday dashboard pulling from the tools you already use, then check the figures against the source once. If it gets them right, the next week is free.

None of those require you to trust Duet with a password, a payment or a signature. They test the part that is hard to get right, which is whether its judgment about what matters matches yours.

Why a directory is watching this launch

SaaSCity is a directory, so the obvious question is why we would spend a morning on an agent app. One reason is that the buyer for a SaaS tool is increasingly an agent. If Duet can compare tools and buy things for a business, then the pages it can read become the pages that get shortlisted. A directory listing is one of the few places where your product's name, category, pricing and description sit in a form a machine can parse without guessing. That is the same argument we make in our guide to submitting a SaaS to directories, and Duet makes it more concrete.

The other reason is simpler. Founders read launch posts like this one to decide what to build next, and every lab shipping a proactive agent is a signal about where the interface is going. If the morning brief becomes normal, the product page that wins is the one an agent can summarize in one line. You can list your product on SaaSCity and see how yours reads.

FAQ

What is Claude Duet?

Claude Duet is a standalone app from Anthropic where Claude works as a proactive agent. It sends a morning brief across your connected apps, takes on tasks by text or voice, keeps working in the background on its own cloud computer, and asks for approval before it sends a message, makes a payment, signs a document or deletes anything. It opened as a research preview on October 21, 2026.

When did Claude Duet launch?

Anthropic opened a research preview of Claude Duet on October 21, 2026, announced on its blog and product page the same day. It is a preview, not a general release: access is a waitlist for Pro and Max subscribers in the United States, and Anthropic describes it as an early version.

How much does Claude Duet cost?

There is no separate price. During the research preview, Claude Duet is included in Claude Pro and Max plans. Pro costs $20 a month, or $17 a month on annual billing at $200 up front, and Max starts at $100 a month. Duet shares your plan's usage limits, so heavy use draws down the same allowance as chat.

Who can use Claude Duet right now?

Pro and Max subscribers in the United States, through the waitlist on claude.com/product/claude-duet. Team and Enterprise plans are not included in the preview, and Anthropic has not published a date for widening access. Anyone outside the US should treat availability as unconfirmed.

Is Claude Duet available in the UK or Europe?

No. Anthropic's announcement starts with the United States and gives no timeline for other countries. The waitlist form on the product page is the only route in, and it does not promise a date. Plans for the UK, the EU and the rest of the world are unannounced as of October 21, 2026.

What model does Claude Duet use?

Claude Opus 5.5. Anthropic says it built Duet around Opus 5.5 because the product depends on judgment: deciding what is urgent, what can wait, which tool to use, and when to interrupt you. The model is the same one Anthropic sells through the API, not a separate Duet-only model.

Does Claude Duet ask for approval before it acts?

Yes, before consequential actions. It shows you what it is about to do and waits before sending a message, making a payment, signing a document or deleting anything. Two actions are off-limits even with approval: trading securities and moving money between accounts. You can also disconnect any app it has access to.

Does Claude Duet see my passwords?

Anthropic says Duet never directly sees your passwords or sign-in tokens, following the credential design used for Claude Managed Agents. It works inside the apps you connect rather than asking you to paste logins into a chat. That is Anthropic's description of the design, not an independent audit.

How is Claude Duet different from Claude?

The regular Claude app waits for you to open a chat. Claude Duet comes to you: a morning brief, periodic monitoring of connected apps, and a message when something needs attention. It carries over your memory, skills and connected apps, runs as a separate app on iOS, Android and claude.ai, and works on its own cloud computer.

Does Claude Duet work while my laptop is closed?

Yes. Duet runs on a virtual machine with its own browser, so it does not need your computer or phone to be on while it works. You get a notification when a task is done or needs your review. Anthropic also says approvals, logins and safety checks run in that same cloud computer.

Does Claude Duet use the same usage limits as Claude?

Yes. Duet shares your plan's usage limits with the rest of Claude, and it uses them differently. Background checks of connected accounts count toward the limits, and larger tasks such as a financial model or a deck need more model turns. Anthropic checks accounts every few hours, rather than nonstop, partly to keep that consumption manageable.

What can Claude Duet connect to?

Anthropic says it can use the more than 6,000 connectors and plugins in the Claude Marketplace, including Gmail, Slack, Google Calendar, Google Drive, Microsoft 365, Notion, QuickBooks, Stripe, Gusto, HubSpot, Docusign, Canva, Shopify and Ramp. It can pay through Stripe's Link with a virtual card, or through Shopify's Shop Pay.

Is Claude Duet the same as duet.so or the claude-duet GitHub project?

No. Duet is a crowded name. duet.so is an unrelated third-party platform that markets cloud agents, and claude-duet on GitHub is an unofficial open-source tool for sharing a Claude Code session between two developers. Anthropic's product is at claude.com/product/claude-duet.

What is Claude Duet bad at?

Anthropic lists the limits itself. It may misread a thread or misjudge what is urgent, it can miss something that arrives between checks, it is stronger at work tasks than personal errands, it cannot show you its browser yet, and it lives in a separate app. Access is also US-only and waitlisted.

Should I build my SaaS on top of Claude Duet?

No. Duet is an end-user app, not an API or a platform to build on. If you want Opus 5.5 in your product, call it through the Claude API. Duet matters to founders as a customer and competitor signal, and as a tool for running the business behind the product.

Sources

• Introducing Claude Duet (Anthropic announcement, October 21, 2026)
• Claude Duet product page and FAQ
• Claude plans and pricing
• Claude Opus 5.5
• Anthropic's framework for trustworthy agents
• Claude Managed Agents engineering post
• Auto mode classifier
• Prompt injection defenses
• Anthropic bug bounty
• What is ChatGPT Dots? (MindStudio, October 1, 2026)
• Meta debuts its Muse AI agent (TechCrunch, September 8, 2026)
• Introducing Grok Bot (xAI, August 11, 2026)

Screenshots were taken from Anthropic's pages on October 21, 2026. Prices and availability can change; check the product page before you plan around them.