On October 21, 2026, Anthropic quietly entered one of the most competitive races in tech: the always-on AI agent. The company opened a research preview of Claude Duet, a standalone app that does not wait for you to type a prompt — it watches your connected apps, briefs you each morning, and keeps working in the background while your phone is off.
The launch made an immediate mark: the search term “claude duet” surged past 20,000 searches in the US within 24 hours, up more than 500% on Google Trends on October 22, 2026. If you run a business, this matters far beyond the hype cycle — Duet is built explicitly for “professionals and small business owners with too much to track,” in Anthropic’s own words, and it previews how your competitors may soon be operating.
Here is what Claude Duet actually is, what it can and cannot do, and what your business should do about it this week.
What Exactly Is Claude Duet?
Claude Duet is Anthropic’s new proactive AI agent, launched as a research preview on October 21, 2026, with access granted through a waitlist. Instead of sitting inside a chat window waiting for instructions, Duet initiates contact: it sends you a short morning brief on the most important items across your connected apps, then checks those apps every few hours and messages you only when something needs your attention.
The product runs on Claude Opus 5.5, Anthropic’s latest flagship model. Anthropic positions Duet as its answer to Meta’s Muse agent and OpenAI’s Dots agent — the same proactive-agent ground that xAI’s Grok Bot and smaller players like Hark, Instinct, and Fo are also chasing. Unlike a conventional chatbot, Duet does its work on its own cloud-based virtual machine with its own browser, so it keeps running after you close your laptop or turn off your phone, and notifies you when a task is finished or needs review.
For now, access is deliberately narrow: it is limited to Claude Pro and Max subscribers in the United States, via the standalone iOS and Android apps and through claude.ai. Team and Enterprise plans are not included, and Anthropic has not announced dates for other countries.
How It Works: The Morning Brief and the Background Checks
The daily rhythm of Claude Duet is deliberately simple:
- The morning brief. Each morning, Duet looks across your email, calendar, and other connected apps, picks out what most needs your attention, and can draft replies for your review.
- Periodic background checks. During the day it checks your connected accounts every few hours and sends a message only when something needs a decision. Anthropic says it deliberately does not check nonstop, because that would consume a plan’s usage limits much faster.
- Work that continues without you. Tasks run on Duet’s own virtual machine in the cloud, so there is nothing to keep open and no device to leave running. You get a notification when work is done or needs your sign-off.
- One continuous thread. You interact with Duet the way you would message a person — typed or voice messages in a single ongoing thread rather than separate conversations. You can create separate channels for different topics that share the same memory, and a dedicated Tasks tab tracks ongoing work. Only you can mark a task as finished.
Crucially, Duet inherits the memory, skills, and connected apps already attached to your Claude account, so there is very little setup. Anthropic says the agent can draw on more than 6,000 connectors and plugins in the Claude Marketplace, including Gmail, Slack, QuickBooks, Stripe, HubSpot, Shopify, Gusto, and DocuSign. The agent does not see passwords or logins when given access to your accounts.
Real Business Tasks It Can Already Handle
Anthropic’s own examples read like a small-business owner’s to-do list — which is exactly who the product is aimed at:
- Invoicing and cash flow. Find unpaid invoices in QuickBooks or Stripe and draft payment reminders. Integration with Stripe’s Link wallet and Shopify’s Shop Pay can facilitate purchases with your approval.
- Payroll. Prepare payroll in Gusto for your review.
- Sales follow-up. Draft follow-up emails from HubSpot contact and deal history.
- Paperwork. Send agreements for signature through DocuSign.
- Ecommerce. Set up a Shopify store.
- Planning and reporting. Build briefs, decks, financial models, sales forecasts, and dashboards.
- Technical work. Even building small web apps — the agent works on tasks in the background on its own machine.
The pattern is clear: Duet is designed to absorb the repetitive operational layer of a business — the invoice chasing, the scheduling, the draft-writing, the report-building — while leaving the decision in human hands.
The Approval Gates: What It Will Never Do Without You
For a business owner, the most important section of Anthropic’s announcement is the one about guardrails. 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.”
Some things are off-limits even with approval: Duet will not trade securities and will not move money between accounts. Anthropic also built a dedicated classifier and safety system prompt that checks each message and guards against sensitive information leaks — the agent declines requests involving credit card numbers and government identification by default, and sensitive information such as health or personal finances is only shared when you have explicitly told Duet to send it.
That approval-first design is the difference between a useful agent and a liability. It means you can give Duet access to your operational systems with a defined boundary: it can prepare and propose; it cannot finalize without you. Mike Krieger, co-lead of Anthropic’s Labs team, said in the launch video that the agent “can have the discernment and judgment around when to be proactive, when to act on your behalf” — and the approval gates are where that judgment is enforced.
Why This Is Bigger Than a Chatbot Upgrade
It is tempting to file Claude Duet under “another chatbot feature.” That would be a mistake. A chatbot answers when asked; a proactive agent notices, prepares, and proposes before you ask. For a business, the difference is structural:
| Reactive chatbots (the old model) | Proactive agents (the Duet model) |
|---|---|
| You open the app and type a question | It briefs you each morning before you ask |
| Work stops when you close the tab | Work continues in the cloud on its own computer |
| You supply all context every time | It inherits your memory, skills, and connected apps |
| One task, one conversation | Ongoing thread, channels per topic, shared memory |
| You remember what to check | It checks apps every few hours and alerts you |
The shift from “AI you consult” to “AI that keeps watch” changes what it means to be organized. Businesses that once competed on who had the best admin hire will increasingly compete on who has the best agent supervision setup — which workflows are delegated, which approvals are required, and how clean the underlying data is.
The Agent Race: Duet, Muse, Dots — and What It Means for You
Anthropic is now the fourth major lab to formalize the always-on agent category, not the first. Meta’s Muse agent targets consumers (with a small-business version reportedly in the works), OpenAI’s Dots integrates professional tools, and xAI’s Grok Bot is in the same race. The specifics of pricing and models differ — Duet is bundled into existing Pro and Max plans during the research preview, using your plan’s normal usage limits — but the direction of travel is identical.
For a business owner, the practical takeaway is not “pick a winner.” It is that proactive agents are now a product category, and every major lab is shipping one. Within a year, the question your competitors will ask is not whether to use an agent, but which agent workflows they trust. Getting comfortable with approval-based delegation now — on your own terms, with your own policies — is the head start.
The Honest Limitations Business Owners Should Know
This is a research preview, and the honest limitations matter more than the demos:
- Availability is narrow. US only, Pro and Max plans only, waitlist only. Team and Enterprise customers are not included, and Europe has no announced date. If you are outside the US, you cannot use it yet — full stop.
- Background activity costs usage. Duet’s background checks and tasks count against your existing Claude plan limits. Heavy delegation eats your allowance.
- Early reviews are mixed. Early reviewer Dan Shipper of Every found Duet’s email drafts weaker than a rival agent’s, and noted its voice capability is “no real voice mode yet, just transcription.”
- It is a preview. Anthropic has said it will share more detail on Duet’s safety design in the coming days. Independent long-term testing does not exist yet, and no business should hand an agent unsupervised control of payments or customer communications in week one.
- Errands are out of scope. Unlike some rivals, Duet cannot handle personal errands or book tickets, and it does not show its browser activity live as it works.
Treat Duet as a signal of where the market is going, not as production infrastructure you can rely on today. That distinction is exactly what separates a business that experiments safely from one that gets burned.
What Your Business Should Do This Week
You do not need waitlist access to prepare for the agent era. The businesses that will benefit most when proactive agents go mainstream are the ones that do this groundwork now:
- Audit your connected data. Duet-style agents inherit your memory and app connections — so a messy inbox, a chaotic calendar, and outdated CRM records become liabilities, not just annoyances. Clean them up before an agent starts reading them.
- Write down your approval policies. Which actions can run on their own, and which always need a human sign-off? Define it now — for your people, and for any agent you connect later.
- Map your repetitive workflows. Invoice reminders, scheduling, follow-ups, report-building, status updates. These are the first tasks agents absorb. Document them once, and automating them becomes a configuration job instead of a project.
- Connect your systems properly. Agents are only as useful as the integrations behind them. If your CRM, accounting, and email still do not talk to each other, that gap is now your biggest bottleneck — not the AI.
- Start with a supervised pilot. When access reaches you, give one agent one supervised workflow, watch the approval gates for a week, and expand only what proves trustworthy.
The era of AI that waits for your prompt is ending. The era of AI that taps you on the shoulder has begun. The businesses that win it will be the ones that are organized, integrated, and disciplined about approvals — long before the agent arrives at their door.
Want help getting your business ready? AI Dev Expert builds AI chatbots, AI calling agents, and AI integrations that connect your CRM, email, and operations — the exact groundwork proactive agents need. We also run marketing automation audits that map which of your workflows should be automated first. Book a free consultation and we will map your highest-ROI agent-ready workflows in one call.
