Claude Duet is here: Anthropic's agent that messages you first

Anthropic launched Claude Duet on October 21, 2026, as a research preview. It is a new, separate app where Claude works as a proactive agent: instead of waiting for you to open a chat, it watches your connected apps, sends you a morning brief and messages you when something needs a decision. It runs on Claude Opus 5.5 and is aimed at professionals and small business owners who have too much to keep track of.

What Claude Duet does

The regular Claude app waits for you to start a conversation. Duet flips that around. Anthropic describes it as "a new app where Claude works as a proactive agent" that "checks with you before it acts on your behalf."

• Morning brief. Each day starts with a short summary of the most important items across your connected apps, often with next steps already drafted.
• Monitoring. Through the day it checks your accounts every few hours and stays quiet unless something needs your attention. You can choose what you want to hear about and how often.
• Tasks of any size. Small jobs like scheduling a meeting, or bigger ones like dashboards, sales or financial forecasts, decks, prototypes and web apps. It tells you when a task is done or needs review.
• One ongoing thread. Everything happens in a single conversation, by text or voice, from quick replies to projects that run for weeks.
• Your Claude setup comes with it. Memory, skills and connected apps carry over from your Claude account, so there is little setup. Duet can also search your past Duet chats and your non-project Claude chats.

Connected apps and business tools

Duet works with the connectors you already use in Claude. Anthropic says it can draw on more than 6,000 connectors and plugins in the Claude Marketplace. For sites without a connector, it uses its own cloud browser.

Anthropic's own examples are very small-business shaped:

For purchases, Duet uses Link (Stripe's wallet) or Shop Pay with a one-time virtual card that works only for that purchase, so your real card number is never shared. Signing in with saved 1Password logins is coming soon.

Approvals and safety

Duet can read, draft and prepare work on its own. Before it sends a message, makes a payment, signs a document or deletes anything, it shows you the exact action and waits for your approval. If you can't answer right away, it gets on with other work and brings the request back later.

Some things are off-limits even if you approve them: trading securities, moving money between accounts and accessing someone else's account.

On security, Anthropic says:

• Duet never sees your passwords or sign-in tokens. Passwords go into a separate form on your device and are sent straight to the browser, not to the chat. The design follows the one used for Claude Managed Agents.
• A new classifier and safety system prompt check messages for sensitive information. The most sensitive data, like credit card numbers and government ID, is rejected. Health and personal finance information is shared only when you explicitly ask.
• Defenses against prompt injection come from Anthropic's browser-use research. Red teams and an outside firm tested the product, and it is covered by Anthropic's public bug bounty.

Availability and pricing

Invites go out in waves, with no guaranteed date. To join, go to the Claude Duet page, enter your name and email, choose your plan and phone type, and describe your work. Use the same email as your Claude account, because a mismatch can block access.

Bigger jobs, like a financial model or a deck, take more model turns and background work, so they use up your limits faster. Pro users in particular should expect to hit limits sooner than they do in normal chat.

Limits Anthropic admits to

Anthropic is unusually upfront about what Duet can't do yet:

• It may "misread a thread or miscalibrate what's urgent," and Anthropic is still tuning when it should message you.
• It checks accounts at set intervals, so something urgent can land between check-ins.
• You can't watch or take over its browser yet.
• It can't make phone calls or send text messages. It can send email from connected accounts, with your approval.
• It can read your work in Claude but can't edit it, and it can't see the projects and files on your computer.
• Personal errands, like booking tickets, are less well supported than work tasks.
• It lives in its own app, separate from the main Claude app, for now.
• If you disconnect an app, Duet loses access, but it keeps its memory of what it already saw.

How it compares

Press coverage frames Duet as Anthropic's answer to OpenAI's Dots and Meta's Muse agent, part of a wider shift toward agents that keep working when you are not chatting. Anthropic hasn't published a side-by-side comparison, and none of these agents has a shared benchmark yet, so judge them on your own workflows. Meta's agent is set to run on Muse Wonder.

Duet's clearest edges are the large connector catalog, the business tools baked into its examples (QuickBooks, Gusto, Stripe, Shopify) and the fact that it reuses your existing Claude memory and skills.

How entrepreneurs can use it

If you run a small business, these are good first jobs to hand Duet once you are in:

1. Cash collection. Connect Stripe or QuickBooks and ask it to flag unpaid invoices every morning with reminder emails ready to approve.
2. Inbox triage. Let the morning brief surface customer emails and Slack threads that have gone quiet.
3. Meeting prep. Ask it to prepare a brief before each client or investor call from your calendar, email and CRM.
4. Recurring admin. Payroll prep in Gusto, contracts in Docusign and follow-ups in HubSpot, all of which stop at an approval step.
5. A daily dashboard. Ask for a short summary of revenue, cash and open orders refreshed before your morning brief.

Start with read-and-draft tasks. Give it the ability to send, pay or sign only once you trust how it judges what matters.

Bottom line

Claude Duet is the first Claude product built to come to you rather than wait for you, and its focus on small-business admin makes it more practical than most "personal agent" pitches. It is also clearly a preview: US-only, invite-only, mobile-first, with usage limits that background work will eat into.

• On Pro or Max in the US? Join the waitlist now. It costs nothing extra, and the launch limit reset softens the usage hit until November 30.
• On Team or Enterprise, or outside the US? Wait. Anthropic says Team and Enterprise are coming later.
• Once you are in: connect a few apps, keep approvals on for everything, and check its drafts carefully for the first few weeks.