Yesterday morning, my inbox had the usual chaos: a dozen unread emails, three meeting invites I never agreed to, and one invoice reminder that was technically my fault. Meanwhile, Anthropic quietly launched the thing that is supposed to end all of that. It’s called Claude Duet, and if the last 24 hours of search traffic are anything to go by, everyone wants to know what it actually is.
Short version: it’s a proactive agent app that pings you when something needs attention, handles tasks while your laptop is off, and can even buy office supplies. It launched as a research preview on October 21, 2026, it’s gated behind a waitlist, and only US subscribers on the Pro and Max plans can get in. Here’s the full breakdown.
What Is Claude Duet?
Claude Duet is Anthropic’s answer to the new arms race in AI agents: assistants that don’t just answer questions but do work on their own. OpenAI launched always-on “Dots” agents last month that chase goals across apps like Slack and Microsoft Teams, and Meta‘s Muse agent pulled in millions of downloads. Duet is Anthropic’s entry in that fight, built around its flagship Claude Opus 5.5 model.
The core idea is overdue: instead of you opening a chat window and typing “summarize my emails,” Duet comes to you with what needs your attention. It checks the apps you’ve connected every few hours, sends you a short morning brief of the important stuff, and messages you during the day only when something actually needs a decision. You talk to it like a person, typed or spoken, in one ongoing thread. It remembers your context, carries over the memory, skills, and connected apps from your existing Claude account, and can even search your past conversations so you don’t have to repeat yourself.
How Claude Duet Actually Works
Here’s the part I find most interesting. Duet doesn’t run on your device at all. It runs on its own virtual machine in the cloud, with its own browser, which means it keeps working after you close the app or shut down your laptop. When a task finishes or needs your review, it sends you a notification.
The day-to-day loop looks like this:
- Morning brief: Duet sends a short rundown of the most important items across your connected apps. Think unread emails that need replies, meetings that conflict, invoices that are overdue.
- Background monitoring: It checks your connected accounts every few hours (deliberately not continuously, more on why in a minute) and messages you when something needs attention.
- Task execution: You ask it to do something, typed or spoken, and it works in the apps you’ve connected, letting you know when it’s done.
- Approval gates: It always asks before doing anything consequential: sending a message, making a payment, signing a document, or deleting a file.
Anthropic says Duet never sees your passwords or logins when you give it account access, and there’s a dedicated classifier plus a safety system prompt that screens each message to guard against sensitive-information leaks.
What Can Claude Duet Do?
Anthropic is pitching Duet squarely at professionals and small-business owners. The tasks it can handle range from the mundane to the genuinely ambitious:
- Scheduling meetings and handling calendar admin
- Building dashboards and financial or sales forecasts
- Drafting follow-up emails and payment reminders
- Prototyping and building web apps
- Checking for unpaid invoices via a Stripe integration
- Buying things: it can purchase business supplies using Stripe’s Link digital wallet or Shopify‘s Shop Pay at online stores. It confirms the item with you, Link asks you to approve the spend, and it pays with a virtual card so your real card number is never shared.
It plugs into more than 6,000 connectors and plugins in the Claude Marketplace, including Gmail and Slack, which is where the morning-brief magic is supposed to come from.
Anthropic itself acknowledges one funny limitation: Duet is “less effective at personal tasks such as booking tickets.” So it can build you a financial model and restock your office paper, but good luck getting it to snag concert seats.
Claude Duet Price and Who Can Get It
During the research preview, Duet costs nothing extra. It uses your existing Claude plan’s usage limits, and the preview is currently limited to Pro and Max subscribers in the United States. Team and Enterprise plans aren’t included yet, though Anthropic says it plans to expand to them later. Access goes out in waves through a waitlist, and Anthropic emails you when your turn comes.
What does that mean in dollars? Claude Pro runs $20 a month ($17 a month if you pay $200 upfront for the year), and Claude Max starts at $100 a month for the 5x-usage tier or $200 a month for the 20x tier. No new subscription to buy, which is a pleasant surprise in an era where every new AI feature arrives with its own price tag.
Here’s the catch hiding in the fine print: background checks count toward your usage limits. Anthropic explicitly warns that Duet’s periodic monitoring can make you hit your limits “sooner than with chat alone.” Larger tasks, like building a financial model or a slide deck, burn through more model turns. So the Pro plan that feels roomy today might feel a lot tighter once an agent is checking your apps behind your back every few hours. (That’s also why it doesn’t check continuously, it would eat your allowance alive.)
Claude Duet vs ChatGPT Dots vs Meta Muse
All three labs have now shipped their version of the always-on agent, and the philosophies differ in ways that matter:
- Claude Duet is the careful one. Approval before sensitive actions, no password visibility, classifier guarding every message, and an explicit admission that it can misjudge urgency. It’s built for professionals who’d rather have the agent ask too often than act too boldly.
- OpenAI’s Dots agents pursue user goals across workplace apps like Slack and Microsoft Teams, positioning more as team workflow automation.
- Meta’s Muse went for the consumer wave first and drew millions of downloads, which is why investors are suddenly very interested in autonomous AI services.
Anthropic’s pitch is that Duet’s judgment is the differentiator: it’s “designed to be resourceful and to decide what needs you, what can wait, and when to check with you first.” Whether that judgment survives contact with real, messy inboxes is the entire ballgame.
The Catches You Should Know About
I’m genuinely excited about this category of product, but let’s be honest about the rough edges, because Anthropic itself listed several:
- It can misjudge urgency. The announcement admits this early version can misread conversations and get the priority wrong. An agent that pings you about a newsletter while your biggest client’s invoice sits unread is worse than no agent at all.
- Background usage eats your plan. Covered above, but it bears repeating: if you’re on Pro and already near your limits, Duet could push you into paying for Max.
- US only, Pro/Max only, waitlist only. Most of the world, and most Claude users, can’t touch it yet.
- It’s a research preview. That label means bugs, behavior changes, and features that may vanish. Don’t rebuild your business workflows around it this week.
- It inherits your Claude context, which is convenient, but it also means any mess in your connected apps becomes mess Duet acts on. Garbage in, confident garbage out.
Should You Join the Claude Duet Waitlist?
If you’re a US-based Pro or Max subscriber who lives in Gmail and Slack, yes, join the waitlist. It costs nothing beyond the plan you already pay for, and being early on an agent that handles your boring admin work is the kind of asymmetric bet I like. Just go in with the right expectations: treat it as a very capable intern, not a co-founder. Give it narrowly defined tasks, review its output, and don’t connect your entire business’s sensitive data on day one.
If you’re not in the US or not on a paid Claude plan, there’s nothing to do but wait. Watch how the preview goes, because the usage-limit economics will decide whether proactive agents become a daily habit or an expensive novelty. If Duet burns through a Pro allowance by lunchtime, Anthropic will have to rethink the pricing math before this goes mainstream.
Final Thoughts
Claude Duet is the clearest sign yet that the chatbot era is ending and the agent era is beginning. The question is no longer “which AI gives the best answers” but “which AI can I trust to act while I’m not looking.” Anthropic is betting that caution wins: ask before acting, never touch passwords, admit when you might be wrong. For professionals handing an AI the keys to their business apps, boring and careful might be exactly the right product. The waitlist is open. I’d get in line.

[…] Opus-class models are paid-only; if you want the agentic, do-things-for-you side, Anthropic’s Claude Duet is still a separate waitlisted product, not part of the normal […]