Claude Duet App: iOS, Android, Web, and the Canada Gap

Google Trends, worldwide, past seven days, read on 23 October 2026, has "claude duet app" at 94, up 100% week over week, with "claude duet canada" up 50%, "how to access claude duet" up 20% and "claude duet ios" up 8% trailing behind it. The product those queries are about is real and is described in some detail. Claude Duet is Anthropic's proactive agent, a research preview on a waitlist for Claude Pro and Max subscribers in the United States, and the product page states that it is powered by Claude Opus 5.5. What the page does not carry is most of what the searches are asking for: a date for any country outside the US, a link to a store listing, or a device requirement. Separating those two lists is the whole job here.

What the app is, in the page's own words

Anthropic's description is that Claude Duet understands what needs your attention, takes on tasks of all sizes, keeps working while you are away, and waits for your ok when it matters. The interaction it describes is a message thread. You hand over work by text or voice, it runs on its own computer in the cloud using apps you connect, and it messages you when the work is finished or when a decision is yours to make.

The FAQ draws the line against the ordinary Claude app in terms of direction rather than capability. The Claude app waits for you to open a chat, and you work on something alongside Claude. Claude Duet comes to you: it tells you each morning what needs your attention, takes tasks you hand it by message, and reaches out when one is done or needs your approval. Anthropic also says that the memory, skills and apps you already have in the Claude app are present from the first message, in one thread, and that Duet works with more than 6,000 connectors and plugins.

Finished work is meant to arrive finished. The page names a doc, a sheet, a deck or a dashboard, and illustrates a morning brief that flags what needs you in your inbox and calendar and drafts replies for you to approve. The same paragraph reaches into small-business chores: finding unpaid invoices and drafting the reminders, preparing payroll for review, getting contracts ready to send.

The setup described is close to none. The FAQ says you message Claude Duet in the app the way you would message a teammate you trust, through text or voice, with no prompts and no configuration. For a preview product behind a waitlist, that is the design claim actually under test: whether an agent with no prompt scaffolding can be handed one sentence and come back with something you would send.

The three surfaces it runs on

The FAQ question "Where can I use it?" has a one-sentence answer, and that sentence is the entire published surface list: Claude Duet is a standalone app for iOS and Android, and on a computer you can open it on the web at claude.ai. Three surfaces, two mobile operating systems, no fourth.

"Standalone app" is doing a lot of work in that sentence for a reader who typed "claude duet app" expecting a download page. The product is an app, and the page calls it one. What the page does not do is point anywhere to download it, which is the subject of a section further down.

Whichever surface you open, the page describes a single history rather than two products. Your conversation is one thread, what you said before is remembered, and the phone and the browser are views onto the same agent with the same connected apps attached.

What the waitlist actually gates

Two things narrow at once. On who can use it, the FAQ says Duet is a research preview on a waitlist for Claude Pro and Max subscribers in the US, with Team and Enterprise plans not included yet. On availability, the page says it is available in Pro and Max plans in the US, and invites you to join the waitlist. So the gate is a plan tier and a country together, and clearing one does not clear the other.

The form on the product page asks for a first name, a last name, the email address on your Claude account, a company name, whether your phone is an iPhone or an Android, and a free-text answer to what you would hand off to Claude Duet first. Phone type is the closest the page comes to acknowledging that the mobile app is what many visitors arrive looking for.

What Anthropic publishes about the queue itself is nothing. There is no position, no batch size, no invitation cadence and no stated order in which countries will be added. Anyone quoting a timeline for an invitation is inferring it, because there is nothing on the page to quote.

How usage is counted during the preview

Duet does not get a separate allowance. During the research preview it shares your plan's usage limits with the rest of Claude, and the FAQ says plainly that it uses those limits differently than a typical conversation. Background checks count toward the same budget, and larger tasks, such as building a financial model or a deck, require more model turns and more background work.

That is a design fact with a practical consequence for anyone planning to run Duet alongside a normal Claude habit on one subscription. A standing morning brief is not free of the usage meter, and the work Duet does while you are away is drawn from the same pool as the work you do in front of it. The page publishes no Duet-specific number, so the only published unit is the general plan limit.

What is published for iOS, Android and Canada: nothing

Take the three rising queries one at a time, because each has the same answer and it is worth stating rather than implying.

"Claude duet canada", up 50% in the window above, has no published answer. The product page names the United States as the only country where Pro and Max subscribers can use Duet, and gives no date, quarter, season or sequence for Canada. The same is true of the United Kingdom and the European Union, which the page does not mention at all. A reader outside the US is not looking at a short delay that has been announced; they are looking at silence.

"Claude duet ios", up 8%, has no published answer in the form people want either. The FAQ confirms that the app exists for iOS and Android, and there is no App Store link and no Play Store link anywhere on the page. Device requirements are unpublished in the same way: no minimum iOS version, no minimum Android version, no handset list, no install size. If a page or a video shows you a store listing for Claude Duet, it did not get that from this product page.

"How to access claude duet", up 20%, resolves to the waitlist and nothing else. The published path is the form described above. There is no published early-access programme, no published partner route, no published enterprise exception and no published way to skip the queue.

The rest of the unpublished list is short but worth having in one place: no waitlist position, no invitation count, no Duet-specific allowance, no per-country rollout order, and no date for Team and Enterprise. Every one of those is a reasonable question, and none of them is answered on the page as read on 23 October 2026.

What Anthropic says is not ready

The FAQ has a section called "What isn't ready yet", and it is unusually specific for a product page. In Anthropic's words, Duet may misread a thread or miscalibrate what is urgent, and it can miss something that arrives between checking in with you. Since the entire premise is an agent that decides when to interrupt you, those two failures are the ones that matter: a wrong urgency call is a bad interruption, and a missed arrival is the opposite failure inside the same mechanism.

Against that, the page describes an approval gate. Duet asks before sensitive operations such as sending messages or paying, and the safety section says approvals, logins and safety checks run in your own cloud computer, so sensitive details stay private. You can see what it has done and what it is doing, and steer it. That is a description of the design, not an audit of it, and it is the only description published.

Access today, and what a straight answer looks like

The published route is short: be a Pro or Max subscriber in the US, then join the waitlist. If your workspace runs on Team or Enterprise, the page says those plans are not included yet, and it describes no separate route for a company that wants the preview before its plan tier is supported.

For everyone else, the honest answer to "how do I access Claude Duet" is that the only published mechanism is a US waitlist with no published timeline, and that any answer claiming otherwise is not coming from Anthropic's page. That is unsatisfying, and it is the accurate summary of what exists as of this reading.

If you need Claude Opus 5.5 before the waitlist clears

Model access and app access are separate things, and separating them is the genuinely useful move for a reader who wants Opus 5.5 capability now without a US Pro subscription. Claude Duet is an agent application. The model beneath it is reached through an API.

OrcaRouter puts more than 200 models behind one OpenAI-compatible endpoint and one key, passes provider list price through with no markup, fails over automatically when a provider has a bad minute, and lets you express routing rules in a routing DSL. Claude Opus 5.5 is $4.00 per million input tokens and $20.00 per million output tokens, with a 1M-token context window and cache reads at $0.20 per million.

The boundary matters as much as the offer. We host models, not Anthropic's apps. An API key does not install Claude Duet, does not add your name to the waitlist, and does not change the country list on that product page. What it gives you is the same model inside your own code, on a schedule you control, which is a different product from a proactive agent that messages you in the morning.

Which brings the original search back around. The app runs on iOS, Android and the web at claude.ai. The waitlist is US-only for Pro and Max. And on Canada, on the store listing and on device requirements, the page is silent, which is itself the published position and the one to plan around.