
Claude Duet Download: What You Can Get Today, and What You Cannot
- OrcaNEWOrca: OrcaCyber Ultra 1.02026-10-17$5.00 / $10.00 per 1M tokens · 13 tok/s
- deepseekNEWDeepSeek: DeepSeek V4.1 Flash 10172026-10-1745Intelligence
- DeepSeekNEWDeepSeek: DeepSeek V4.1 Flash2026-10-1740Intelligence
- OrcaNEWOrca: OrcaCyber Zero 1.52026-10-10$3.00 / $7.50 per 1M tokens · 154 tok/s
- openaiOpenAI: GPT-6.1 Sol2026-09-2952Intelligence
- anthropicAnthropic: Claude Sonnet 5.52026-09-2856Intelligence
- typesafeTypeSafe: Jev 1.132026-09-24$0.04 / $0.00 per 1M tokens · 983 tok/s
- OpenAIOpenAI: GPT-6 Luna2026-09-2238Intelligence
- OpenAIOpenAI: GPT-6 Sol2026-09-2248Intelligence
- AnthropicAnthropic: Claude Opus 5.52026-09-2258Intelligence
- xAIGrok 4.72026-09-2146Intelligence
- OrcaOrca: OrcaCyber Zero 1.02026-09-17$3.00 / $7.50 per 1M tokens · 95 tok/s
- OrcaOrca: OrcaVerify Text 1.02026-09-16$2.00 / $0.00 per 1M tokens · 964 tok/s
- OpenAIOpenAI: GPT-6 Astra2026-09-0453Intelligence77Coding
- GoogleGoogle: Gemini 3.8 Flash2026-09-0241Intelligence76Coding
- AlibabaQwen: Qwen3.8 Max (0902)2026-09-0245Intelligence76Coding
- AnthropicAnthropic: Claude Fable 5.12026-09-0153Intelligence82Coding
- TencentTencent: Hy4 preview2026-08-28$0.83 / $2.50 per 1M tokens · 47 tok/s
- AlibabaQwen: Qwen3.8 Flash2026-08-26$0.15 / $0.47 per 1M tokens · 100 tok/s
- z-aiZ.ai: GLM 5.3 Flash2026-08-2642Intelligence72Coding
Search for a Claude Duet download and you will land in one of three places: Anthropic's own announcement, a news piece repeating it, or a page offering an APK, an invite code or an "early access" upgrade. Only the first is a route to the software. As of 22 October 2026, the day after Anthropic opened its research preview, Claude Duet, the proactive agent Anthropic designed around Claude Opus 5.5, is reached through a waitlist rather than a store shelf. There is nothing to sideload while you wait, and no third party has a faster path in than the form Anthropic published itself.
What does exist is a real product with a defined shape. Anthropic describes Claude Duet as a standalone app on iOS and Android, and on claude.ai — the same Claude account, a different surface. It is a proactive agent rather than a chat window: it comes to you with what needs attention instead of waiting to be asked. The preview is included in Pro and Max plans, shares their usage limits, and starts with subscribers in the United States. Anthropic's instruction to anyone outside that group is to join the waitlist and wait for an email.
The app exists. The download button does not, yet.
The distinction matters because search results blur it. Anthropic says the app is available as a standalone app on iOS and Android and on claude.ai, which is a statement about the product's form, not a statement that either app store currently lists it for you. Whether a listing is live, in review, or gated by country and plan is not something an announcement page says, and we are not going to guess at it here. The one mechanism Anthropic names for access is the waitlist, and the one promise it makes is that it will email as soon as Claude Duet is ready for you.
So the useful thing to understand first is that the waitlist is not a marketing form attached to a product you can already install. It is the distribution channel. For this preview, access is allocated by Anthropic, to Pro and Max subscribers in the United States, and there is no published timetable for how quickly that allocation moves. Anyone who tells you their page shortens the wait is describing something other than Anthropic's process.

What you are waiting for, described without the launch copy
The thing on the other side of the waitlist is narrower than the announcement's enthusiasm suggests, and knowing its shape is useful before you spend a plan on it. Claude Duet sends one short brief each morning summarising the most important items across the apps you have connected. Through the rest of the day it stays quiet unless something needs you: an unpaid invoice, a calendar collision, a thread that has stopped waiting. It runs on a virtual machine with its own browser, so the work continues while your laptop is shut and your phone is in your pocket, and it notifies you when a task is finished or needs review.
Everything it does on your behalf passes through a gate. Before it sends a message, makes a payment, signs a document or deletes anything, it shows you the exact action and waits. A shorter list sits behind that one: trading securities and moving money between accounts are off-limits even when you approve them, and credit card numbers and government identification are refused at the message layer rather than at the approval layer. If those refusals sound like missing features, this is probably the wrong product for you, and finding that out before an invitation arrives is cheaper than finding it out after.
The plan the preview actually requires
Duet is bundled into Pro and Max rather than sold as its own subscription, which makes the plan question unusually concrete. Anthropic's pricing page currently lists four consumer and team tiers. Pro runs $17 per month on annual billing, paid $200 up front, or $20 month to month. Max comes in two sizes, 5x at $100 per month and 20x at $200 per month. Team is priced per seat on a separate track. The preview starts with the first two, in the United States, and it uses the same usage limits as the rest of Claude rather than a separate Duet allowance.
That last clause is the one people underestimate. Background checks count toward those limits, and Anthropic says plainly that larger jobs — building a financial model, assembling a deck — require more model turns and background work. An agent that checks your accounts every few hours is spending your quota while you sleep. If you are on Pro and already near the ceiling in a normal month, Duet will not sit politely on top of that usage; it will compete with it, and the app itself warns that checks run every few hours rather than continuously precisely because nonstop polling would drain the plan faster. The cadence is a budget decision wearing the costume of a design decision, which is worth knowing before you decide which tier you need.

If a page offers to sell you access
There is no legitimate resale of Claude Duet access, because access is not a licence Anthropic issues to anyone else. It attaches to a Pro or Max subscription on your own account and is released by Anthropic's waitlist. Any page charging for an invite code, an "activation", a "premium queue position" or a download of the app is selling something Anthropic did not make, and the file you receive is at best useless and at worst not what it claims to be. The same applies to mirrors of an installer: a phone app delivered outside the store it belongs to is a phone app someone else has had their hands on. The only address to check is Anthropic's own, and the only thing to hand over is your email on the waitlist form.
What to have ready before the email arrives
Setup is deliberately light, and that is the reason a little preparation goes further than usual. Claude Duet carries over your memory, skills and connected apps from your Claude account, which means the agent inherits your existing context rather than starting from a blank profile. The corollary is that the decisions you have already made about which tools Claude can reach will follow you in. If your account has accumulated application access over a year of use, Duet inherits all of it on day one, and the moment to review that list is now, not after an agent with a browser starts working through it unattended.
Two habits are worth forming while you wait. The first is deciding what you want to be told about. Anthropic says you can tell Duet what to hear about and how often, and that its judgment about when to message is still being tuned, so a written list of what genuinely counts as urgent for your work will be more useful than reacting to the first week of notifications. The second is deciding what you would never want it to send, because that is the part it cannot decide for you. Health information and personal finances pass the classifier only when you have explicitly instructed it. Everything stricter than that does not pass at all. Knowing where your own line sits before the tool asks is the difference between approving actions quickly and reading every draft twice.

There is also the practical matter of money moving. Duet can buy things for your business, but it does so through Stripe's Link wallet or Shopify's Shop Pay, and Link pays with a virtual card so your real number is never handed to the merchant. Signing in to sites with credentials saved in 1Password is listed as coming soon rather than shipped. If purchasing is part of what you want from the agent, setting up the wallet and deciding its limit is an afternoon's work you can do before the invitation, and it removes the awkward first session in which you are configuring payment rails while an agent waits.
One more thing to expect: it is a second app
Claude Duet lives in its own app, separate from the main Claude app, and Anthropic says what it learns from early users will shape how the two eventually work together. For now that means another icon, another notification stream, and a second place where context lives. Anyone who has watched a company split one assistant into three surfaces knows how this goes: the split is real, the integration is a roadmap item, and the burden of checking both lands on you for as long as the split lasts. Anthropic is straightforward that the preview is early, that it may misread a thread or miscalibrate what is urgent, that personal errands are weaker than work tasks, and that it cannot yet show you its browser. Those are the sentences to re-read when the first week feels underwhelming.
What runs underneath, if you would rather build than wait
Anthropic designed Duet around Claude Opus 5.5 and built it to judge what needs you, what can wait, and when to check first. That model is on our catalogue, and the reason it is worth naming here is that not everyone reading a download page wants a consumer app. If what you actually want is a proactive agent over your own tools, with your own prompts and your own schedule, the model is the part you can have today: Claude Opus 5.5 at $4.00 per million input tokens and $20.00 per million output tokens, a 1M-token context window, cache reads at $0.20 per million, served through one OpenAI-compatible endpoint and one key that also reaches more than 200 other models. We pass provider list price through with no markup, so the price you see is the vendor's, not ours with a layer on top.
To be exact about what that is and is not: we host the model, not the app. Claude Duet is Anthropic's consumer product, distributed by Anthropic to its own subscribers, and nothing on a model catalogue gives you the scheduling, the morning brief or the approval interface. What it gives you is the same reasoning engine those things are built on, callable from your own code, which is the difference between waiting for an invitation and shipping something this month.
What to do this week
Join the waitlist if you are on Pro or Max in the United States, and do nothing else that costs money. Then spend twenty minutes on the two things the invitation cannot do for you: pruning the apps your account has connected, and writing down what you actually want to be interrupted about. Check where your plan's usage sits, because an agent that runs background checks is a new consumer of the same allowance, and the first month is a poor time to discover the ceiling. And if a page offers to speed any of this up for a fee, close it — the only thing that moves you up the list is Anthropic deciding it is your turn, and the email is the only notification that will tell you so.
