Claude Duet Is Anthropic's First Personal Agent: What It Does, Who Gets It, and the Usage-Limits Catch

Every AI assistant until now has waited for you to open it. On October 21, 2026, Anthropic flipped that around: the company announced Claude Duet, a new app where Claude works as a proactive agent that messages you first — a morning brief of what needs your attention, tasks that keep running while you step away, and an explicit ask for approval before it does anything that matters. The company describes it in its own words: "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."

That puts Anthropic in direct competition with Meta's Muse personal agent and OpenAI's always-on Dots agents, both of which launched in September. But the details of who gets Duet, what it costs, and what it quietly consumes make this launch worth reading closely before you join the waitlist.

The quick answer

What it is: A standalone app (iOS, Android, claude.ai) where Claude Opus 5.5 acts as a proactive personal agent — surfacing tasks, completing them in the background, and asking for approval on consequential actions.
Who gets it: US subscribers on Claude Pro or Max plans, via a waitlisted research preview. Team and Enterprise plans are not included yet, and Anthropic has named no date for international expansion.
What it costs: No extra charge during the preview — but it shares your existing plan's usage limits, and Anthropic warns those background checks can make you hit the cap "sooner than with chat alone."
The model: Claude Opus 5.5, Anthropic's top reasoning model, which launched in September at what the company said was 40% lower cost than Opus 5.
Safety model: Duet shows you exactly what it plans to do before it sends a message, makes a payment, signs a document, or deletes a file — and it never sees your passwords.

What Claude Duet actually does

The centerpiece is the morning brief. Each day, Duet sends a short summary of the most important items across your connected apps, then checks those accounts every few hours and messages you only when something genuinely needs attention. Anthropic said it deliberately doesn't check nonstop, because continuous polling would burn through a plan's usage limits much faster — a trade-off it states plainly on the product page.

Tasks you hand it can be small or substantial: scheduling meetings on the simple end; building dashboards that pull from dozens of tools, creating financial or sales forecasts, prototyping, and building web apps on the ambitious end. The agent runs on its own virtual machine with its own browser, so it keeps working after you close the app, and it sends a notification when a task is complete or needs your review.

Setup is minimal by design. Duet inherits the memory, skills, and connected apps from your existing Claude account, and it can search your past conversations so you don't repeat yourself. Interactions happen in one ongoing thread instead of scattered conversations, and you can message it by typing or by voice — "the way you'd message a person," as Anthropic puts it.

The connector library is the real product

Anthropic says Duet works with more than 6,000 connectors and plugins in the Claude Marketplace. The logos on the product page read like a small business's entire stack: Stripe, 1Password, Shopify, Gmail, Google Calendar, Slack, Intuit QuickBooks, DocuSign, Google Drive, HubSpot, Gusto, Microsoft 365, Notion, Canva, and Ramp.

The examples Anthropic gives are all work-focused: checking QuickBooks or Stripe for unpaid invoices and drafting payment reminders, preparing and running payroll in Gusto, drafting follow-up emails from HubSpot contact and deal history, sending agreements for signature through DocuSign, and setting up a Shopify store. Purchases go through Stripe's Link digital wallet and Shopify's Shop Pay — with your approval, every time.

That's the clearest signal about who Duet is built for: professionals and small business owners "with too much to track," in Anthropic's words, rather than general consumers. It also explains the US-first, Pro-and-Max-first rollout — these are the customers whose tool stacks and usage limits make an agent like this viable.

Pricing: included, but the limits are the fine print

During the research preview, Claude Duet is included in Claude Pro and Max subscriptions at no additional charge. The published plan prices, verified October 2026:

Claude Pro: $20 per month, or $17 per month billed annually ($200 up front).
Claude Max: from $100 per month, with a choice of 5x or 20x Pro's usage allowance.

The fine print matters more than the headline. Duet uses your existing Claude plan limits, and everything it does in the background counts against them — the morning brief's checks, the periodic app monitoring, all of it. Anthropic's own phrasing: you may reach your limits "sooner than with chat alone." For a Max subscriber with a large allowance, that may be invisible; for a Pro user with a modest one, a Duet left monitoring three busy accounts could quietly eat the chat budget.

Three gaps stand out in what's been disclosed: there is no confirmed standalone price for Duet if Anthropic ever unbundles it; Team and Enterprise customers — the segment most likely to use an agent at scale — have no access yet; and there is no date for availability outside the US. The company says it plans to expand to Team and Enterprise eventually. Until then, the preview is effectively a retention product for individual subscribers, not an enterprise offering — and buyers should not assume it carries the audit logs and custom retention controls that Anthropic lists under Enterprise.

The safety model: ask first, always

Where Anthropic draws its line is the most distinctive part of the launch. Before Duet sends a message, makes a payment, signs a document, or deletes a file, it shows you exactly what it's about to do and waits for your approval. Trading securities and moving money between accounts are off-limits entirely — even with approval. A new classifier is designed to block requests to share credit card numbers, government IDs, and other sensitive details, and sensitive information like health or personal finances goes out only when the user has explicitly told Duet to send it.

On the plumbing side, Duet never directly sees your passwords when given account access, and approvals, logins, and safety checks run in the user's own cloud computer so sensitive details stay private. Sign-ins using logins saved in 1Password are coming soon, building on the 1Password–Claude login feature from July.

Crucially, Anthropic states the caveats in the open rather than burying them. "Claude Duet is a research preview. It may misread a thread or miscalibrate what's urgent, and it can miss something that arrives between checking in with you." Reuters also notes the company admits the agent remains less effective at personal tasks such as booking tickets. That's an honest framing: the hard problem for a proactive agent isn't the tasks, it's the judgment — when to interrupt, and whether to trust it with your inbox and your money.

The race it's entering

Anthropic is late to a party that started last month. OpenAI launched always-on Dots agents on September 29, 2026, for ChatGPT Pro and Business Premium users, pursuing user goals across apps like Slack and Microsoft Teams. Meta's Muse personal agent drew millions of downloads after its September launch and already shops, books, and sends emails for users. SpaceXSI's Grok Bot is in the same fight. All four give their agents a cloud computer to keep working in the background — the architecture is now the industry standard; the differences are in philosophy.

Anthropic's bet is the conservative one: a prepare-and-draft agent that pauses for human approval before anything consequential, versus Meta's more autonomous agent that opens browsers, fills forms, and negotiates on a user's behalf. In a market where autonomous agents have already produced embarrassing mishaps — in one incident flagged during launch coverage, Meta's Muse reportedly handed a stranger a user's home address while selling his keyboard — "asks before it acts" is less a feature list and more a thesis. Whether users prefer the cautious agent or the one that just gets things done will be the interesting consumer question of the next quarter.

What to watch next

Anthropic published no benchmark results, no task-level cost estimates, and no quantified productivity claims for Duet — everything in the announcement is descriptive, which means independent testing will have to establish whether the Opus 5.5 agent actually outperforms its rivals on real tasks. Watch for three things: how fast the waitlist moves (invitations go out in waves), whether Anthropic closes the Team/Enterprise gap, and whether the approval-first model proves to be a moat or a handicap once millions of users are comparing agents side by side. Meta has already tested higher-priced tiers for Muse; if Duet catches on, a standalone price is likely the next lever Anthropic pulls.

Frequently asked questions

What is Claude Duet?

Claude Duet is a proactive AI agent app from Anthropic, currently in research preview. It identifies tasks needing your attention across your connected apps, completes work on its own cloud computer, keeps working while you're away, and asks for your approval before taking consequential actions like sending messages or making payments.

Who can access Claude Duet right now?

Access is limited to Claude Pro and Max subscribers in the United States, via a waitlist with invitations going out in waves. Team and Enterprise customers are not included yet, and Anthropic has not announced dates for international expansion.

Does Claude Duet cost extra?

No — during the research preview it's included in Pro ($20/month, or $17/month billed annually) and Max (from $100/month) subscriptions at no extra charge. But it shares your plan's usage limits, and its background checks count toward them, so you may hit your cap sooner than with chat alone.

How is Claude Duet different from Meta Muse or OpenAI Dots?

All three are proactive agents that work in the background on cloud computers. The difference is autonomy: Duet is built to prepare and draft work, then pause for your approval before sending, paying, signing, or deleting. Muse and Dots are designed to act more independently toward your goals. Anthropic's framing is that the approval step is the product.

Is Claude Duet safe to connect to my business tools?

Anthropic's safety design has the right ingredients — it never sees your passwords, a dedicated classifier blocks sensitive leaks like credit card numbers, and securities trading and money transfers are forbidden outright. But the company itself warns the preview "may misread a thread or miscalibrate what's urgent." Connect read-only workflows first, keep approvals on for everything, and don't hand it financial actions until you've seen how it behaves on your own data.

Key takeaways

Anthropic's Claude Duet (Oct 21, 2026) is the company's first proactive personal agent — it briefs you each morning and works on its own cloud computer while you're away.
It runs on Claude Opus 5.5 and plugs into more than 6,000 app connectors, with small-business workflows (invoices, payroll, contracts, stores) as the headline use cases.
US-only, waitlisted research preview for Pro ($20/mo) and Max (from $100/mo) subscribers — no extra fee, but background activity draws from your existing usage limits.
The approval-first safety model is the differentiator: Duet shows you exactly what it plans to do before sending, paying, signing, or deleting, and some actions are banned outright.
The open questions are judgment, not features — Anthropic admits it can misread threads and misjudge urgency, so independent testing will decide whether the cautious agent wins.

Claude Duet is the least dramatic agent launch of the year and possibly the most interesting one — because the feature it's really selling isn't what it does while you watch, but what it does while you don't. Whether that's a comfort or a liability depends on how much you trust an AI with your inbox at 3 a.m. As of October 22, 2026, Anthropic is asking you to decide that on its approval dialog, one task at a time.

Sources

10 min read

4 views

1h ago

What are you looking for?

Share This Page