What Is Claude Duet? Anthropic's Proactive Personal Agent, Who Gets It, What It Costs, What It Can and Can't Do, and How It Compares to Meta Muse and ChatGPT's Dots

Update, October 22: we're in. Our editor's waitlist invitation arrived Wednesday night, so this now carries a first day of actual use, with the caveat that one day is an impression, not our errands test. Two things stand out. First, it is very proactive, arguably too proactive: the check-ins arrive more often than a morning brief and a quiet day would suggest, and the instinct to surface things borders on nagging, which is a tuning problem Anthropic has already flagged ('it can still misjudge what's urgent') rather than a design flaw. Second, the memory carryover is real and immediate. Because Duet starts from the memory, skills, and connectors of the main Claude app, it knew a great deal about our editor's work from the first message, which made the first brief useful in a way a cold-start agent can't be, and which anyone who hasn't reviewed what Claude remembers about them should check before switching it on (Settings > Memory in the main app). Net: pleasant, useful, and a little much. The launch post, meanwhile, has gone from 1.1 million views on Wednesday evening to 7.5 million and 49,800 likes by Thursday morning, which makes it the biggest Anthropic consumer announcement of the year by that measure. Claude's rating in the personal agents ranking still holds at 4.1 until Duet has run the same reservation, email, purchase, and scheduling errands as the other five; that happens this week.

What Anthropic says it is

From the product page: 'Claude Duet is a proactive agent that understands what needs your attention. It takes on tasks of all sizes, keeps working while you're away, and waits for your ok when it matters.' The announcement, dated October 21 and posted on claude.com rather than Anthropic's newsroom, explains the difference from the Claude you already use: 'The Claude app waits for you to open a chat… Claude Duet comes to you.' The @claudeai post at 12:09 pm PT put it in one line, '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,' and a follow-up added the caveat Anthropic repeats everywhere: 'It can still misjudge what's urgent.' Three things carry over from the main app: 'Your memory, skills, and apps from the Claude app are there from the first message, all in one thread.' One thing doesn't: 'For now, Claude Duet lives in its own app, separate from the main Claude app.'

• Morning brief and check-ins. 'Claude Duet briefs you every morning, drafts next steps, then goes quiet until something needs you.' It checks connected accounts 'every few hours, because checking nonstop would use up your plan's limits much faster.'
• Tasks of all sizes. Set up a meeting, build a dashboard, chase an invoice, draft the deck. The app has a Tasks tab (only you can mark one complete), a Make tab for finished files, and separate channels per topic that share one memory.
• Text or voice, one thread. 'Message Claude Duet in the app like a teammate you trust, through text or voice with no prompts or setup.'
• 6,000-plus connectors. Named at launch: Gmail, Slack, QuickBooks, Stripe, Gusto, LinkedIn, DocuSign, HubSpot, Canva, Shopify, plus the Claude Marketplace catalog. Microsoft 365 'isn't available yet'; Slack 'works with limits.'
• Payments. It can buy for your business through Stripe's Link wallet and Shop Pay, each purchase on a one-time virtual card you approve in the wallet app; Anthropic says Claude 'isn't given your full credit card number.' US payment method required; no recurring payments. Saved 1Password logins are 'coming soon.'
• Approvals. 'Before it sends a message, makes a payment, signs a document, or deletes anything, it shows you exactly what it's about to do and waits for your approval.' Off-limits even if you say yes: 'trading securities, moving money between accounts, or accessing someone else's account.'

Who gets it, and what it costs

Three requirements from the help center: a Pro or Max plan, a Claude account in the US, and an invitation from the waitlist, which 'opens in waves.' You join from the product page with your name, email, plan, and phone type; the invite arrives by email and is tied to that address. Team and Enterprise are 'not included yet' with a promise to 'bring it to those plans later.' The price is the plan you already pay: 'During the research preview, Claude Duet is included with the Pro and Max plans at no extra cost.' Pro is $17 a month billed annually or $20 monthly; Max is $100 or $200 depending on the usage multiplier. The catch is in the usage: Duet 'uses your plan's usage limits, including the five-hour session and weekly limits,' and 'work it does in the background, like the morning brief and check-ins, also counts, so you may reach your limits sooner than with chat alone.' Pro users should expect that to bite. Anthropic hands admitted users 'one full usage limit reset' as a welcome gift, which tells you it expects the problem.

What isn't ready, in Anthropic's words

• 'It may misread a thread or miscalibrate what's urgent, and it can miss something that arrives between checking in with you.'
• 'Personal errands, such as booking tickets, aren't quite as strong.' This is a work agent first.
• 'It can't yet show you its browser.' You see what it has done and is doing, but you can't watch or take over the session.
• 'You can't choose a model. Claude Duet runs on Claude Opus 5.5.'
• It 'doesn't make phone calls or send text messages.' It messages you inside its own app.
• No Team or Enterprise; no Microsoft 365; US only; iOS, Android, and web only; no desktop app.
• Disconnecting an app stops future access, 'but memory of past information from earlier conversations is kept.' There is no documented Duet off switch beyond disconnecting apps and payment methods, and the help articles say nothing about training or retention specific to Duet; Claude's standard consumer settings apply.

Safety: agent security, not wellbeing

Because the name invited the wrong guess, it's worth being explicit: there is no companionship, emotional-support, or relationship language anywhere in Duet's pages. The safety section is about an agent holding your credentials and your wallet. Credentials follow the Managed Agents design 'so it never directly sees your passwords or sign-in tokens'; you type passwords into a separate form on your own device. Actions are 'approved by our auto mode classifier,' prompt-injection defenses come from Anthropic's browser-use research, and the system was 'penetration tested' by red teams and an outside firm and sits under the public bug bounty. New for Duet is 'a dedicated classifier and safety system prompt' that refuses to share 'the strictest tier of information, including credit card numbers and government identification,' and passes health or personal-finance details 'only when you've explicitly instructed Claude Duet to send it.' Anthropic says more safety detail is coming 'in the coming days.' Read against our Muse permissions guide, the model is the same one-time-card, approve-in-wallet pattern Muse uses, with the same legal consequence: a Link purchase you approve is yours.

Claude Duet vs Muse, dots, Grok Bot, Instinct, and Gemini Spark

Facts from each vendor's pages as of October 21; where a price or rule is only in press reports, it's marked. Our ranks are from the personal AI agents ranking, where Duet does not yet have a score because we're on the waitlist like everyone else.

The shape of the field is now clear. Meta went free and public; OpenAI gated its agent behind the $100 tier; Anthropic split the difference, $20 Pro gets in but by waitlist and with shared limits. Every one of them uses a cloud computer, a morning-brief pattern, and Stripe's one-time-card checkout; the differences are reach (Instinct texts you, dots is learning to, Duet and Muse message in-app), personal-errand strength (Muse and Instinct lead, Anthropic admits Duet lags), and defaults (Anthropic's are the most conservative, which is why Claude sits fifth in a ranking that scores what agents do).

Our read

Duet fixes the two things we marked against Claude as an agent: it comes to you, and it can spend. What it doesn't fix yet is the thing that decides this category, which is whether it's better at the errands than Muse and Instinct, and Anthropic is unusually candid that it isn't, for now. The business framing is smart. Muse is pitched at everyone; Duet is pitched at the person who has QuickBooks, Gusto, and a Shopify store open in three tabs, and for that person a Claude that chases invoices and drafts the renewal redline is worth the Pro price even before you get to errands. The usage-limit overlap is the real cost: an always-on agent eating a $20 plan's five-hour windows will push people toward Max, which may be the point. We've joined the waitlist. Claude's score in the personal agents ranking stays at 4.1 until we've run Duet through the same errands as the rest, and its entry now records what Duet changes. If you're deciding today: Muse if you want an agent now and free, Duet if you're a Claude Pro user in the US who runs a business and can wait for the wave. Related: the Claude review and Is it safe to connect Gmail and your bank to Muse?