Learning how to use Claude Duet starts with one question: what should an AI agent handle on its own, and what should it ask you about first? Anthropic announced Claude Duet on October 21, 2026, as a proactive AI agent for professionals and small business owners who have more work than they can track. It sends a morning brief, takes on tasks you hand it, keeps working when you step away, and checks with you before it acts on your behalf.
Claude Duet is a research preview, and access is limited. It opened by waitlist to US Pro and Max subscribers. This guide explains what Claude Duet is, who can access it today, how it works, how to set it up, a practical workflow, realistic use cases, pricing, safety safeguards, and current limitations.
Claude Duet Explained: Key Features and Takeaways
- Proactive by design: Claude Duet sends a short morning brief and messages you when something needs attention.
- Works in the background: It runs on its own cloud computer with a browser, so your laptop or phone does not need to stay on.
- Uses your existing Claude context: Your memory, skills, and connected apps carry over, and you message it by text or voice in one thread.
- Asks before consequential actions: It shows you what it plans to do before sending messages, making payments, signing documents, or deleting anything.
- Limited access: It is a research preview on a waitlist for US Pro and Max subscribers.
- Key limitation: It can misjudge what is urgent or miss something that arrives between its checks, and its personal-errand features are weaker than its work features.
Claude Duet is Anthropic’s proactive AI agent that takes on tasks you delegate, works in its own cloud environment, and messages you when work is finished or needs your decision.
What Is Claude Duet and How Does It Work?
The main difference between Claude Duet and a conventional chatbot is who starts the conversation. A standard chat waits for you to open it and ask. Anthropic describes Claude Duet as something that comes to you. Each morning it sends a brief covering the most important items across your connected apps, then it monitors those apps and speaks up only when something needs you.
You delegate work by describing it in a message. Claude Duet works out which apps to use and what steps to take. Anthropic says it has its own virtual machine with a browser, so a multi-step task can continue after you close the app. When the task is done or needs your review, you get a notification.
Claude Duet also checks your connected accounts every few hours rather than constantly. Anthropic says nonstop checking would use up your plan’s limits much faster. The practical effect is that urgent items can wait up to a few hours to surface, which matters when you set expectations for it.
Anthropic says Claude Duet is built around Claude Opus 5.5. Because it carries over your Claude memory and skills, it starts with the context you have already built rather than a blank slate.
Claude Duet vs. Claude: What Is the Difference?
Claude Duet is a separate app during the research preview, not a mode inside the main Claude app. Anthropic says what the team learns from early users will shape how the two work together.
| Feature | Standard Claude | Claude Duet |
| Interaction model | You open a chat and direct the work | Duet sends a morning brief and messages you when needed |
| Task delegation | Works within the conversation you start | Takes multi-step tasks and works in the background |
| Background execution | Work happens while you are in the chat | Continues on a cloud virtual machine when you step away |
| Notifications | Responds when you return | Notifies you when work finishes or needs a decision |
| Context | Memory, skills, and connectors from your account | Same account context, in one ongoing thread |
| Approvals | You direct each step | Asks before sending, paying, signing, or deleting |
| Availability | Available across Claude plans | Research preview, US Pro and Max by waitlist |
Standard Claude already handles a lot of agentic work, including tasks with connected tools. Duet’s difference is that it is designed to start work on your behalf and reach out to you, not wait for a prompt.
How to Get Access to Claude Duet
- Check your eligibility. Anthropic says Claude Duet is included in Pro and Max plans for US users. Team and Enterprise plans are not included yet.
- Join the waitlist. Go to the official product page at claude.com/product/claude-duet and join the waitlist.
- Wait for Anthropic’s email. Anthropic says it will email you when Claude Duet is ready for you. Joining the waitlist does not guarantee immediate access.
- Open the app after you are admitted. Use the Claude Duet app on iOS or Android, or open it on claude.ai, and sign in with your Claude account.
Claude Duet is not currently available outside the United States according to Anthropic’s official pages. Anthropic has not announced a timeline for other regions, so check the product page for updates before planning around it.
How to Set Up Claude Duet and Connect Your Apps
Setup requires little work because Claude Duet uses your existing Claude account. Your memory, skills, and connected apps come with you, so you start with the context you already have.
Anthropic says Claude Duet can use more than 6,000 connectors and plugins in the Claude Marketplace. Its product page shows examples including Gmail, Google Calendar, Slack, Google Drive, Microsoft 365, Notion, Intuit QuickBooks, HubSpot, Gusto, DocuSign, Stripe, Shopify, Canva, and Ramp. Some capabilities are still coming. Anthropic lists 1Password sign-in and a richer Shopify storefront editor as coming soon, so do not plan around them yet.
Before you connect an app, review what it can access. Anthropic says you can disconnect any app Claude Duet has access to. Anthropic also says Claude Duet never directly sees your passwords or sign-in tokens, which is why connections should be reviewed rather than granted without thought.
Start with one or two apps tied to a specific workflow, such as email and calendar, then expand once you have seen how Duet handles them.
How to Use Claude Duet: Step-by-Step Guide
Step 1. Choose a Clear Task
Pick work with a defined outcome. Good first tasks include preparing a morning briefing, reviewing overdue invoices, or organizing information for tomorrow’s meetings. Avoid open-ended goals like “run my business” until you understand how Duet behaves.
Step 2. Describe the Desired Result
State the outcome, the sources to use, any deadline, and what should wait for your review. Here is an illustrative prompt. It is an example written for this guide, not an official Anthropic template:
“Review my connected email and calendar. Prepare a briefing for tomorrow’s meetings, identify scheduling conflicts, summarize relevant messages, and list decisions that need my attention. Do not send messages or change calendar events without asking me.”
The last sentence matters. Stating the boundaries up front helps Duet know where to stop and ask.
Step 3. Let Claude Duet Work
Once you send the task, Duet works out the apps and steps it needs and carries them out in the background. It does not require your device to stay on. Expect multi-step tasks to take time, and expect them to consume usage, which is covered below.
Step 4. Respond to Notifications
Duet messages you when a task is complete, when it needs a decision, or when something in your accounts needs attention. Read each request before responding. The purpose of the approval step is to let you catch mistakes, so treat each notification as a checkpoint rather than a formality.
Step 5. Review and Refine the Result
Duet returns finished work such as a document, spreadsheet, or deck for your review. Check names, dates, figures, and recipients against the original sources. If something is off, reply in the same thread and specify the correction, such as “The second invoice amount is wrong. Use the figure in the Stripe record.” Clear feedback produces better revisions.
Step 6. Approve Consequential Actions Carefully
Before Duet sends a message, makes a payment, signs a document, or deletes anything, it shows you exactly what it plans to do and waits for approval. Verify recipients, amounts, and document contents before you approve. Anthropic also states that some actions are off-limits even with approval, including trading securities and moving money between accounts.
Purchases work through Link, Stripe’s digital wallet. Duet confirms the item with you, Link asks you to approve the spend, and payment uses a virtual card so your real card number is not shared.
Before delegating any task, define the result, the boundaries, and how you will check the output.
Practical Claude Duet Use Cases for Work
| Use case | Example task | Human review needed |
| Email and calendar | Flag unanswered messages and draft replies | Check tone, recipients, and any commitments before sending |
| Meeting preparation | Build a brief for upcoming calls from calendar and email | Verify facts and figures against the original sources |
| Invoices and payments | Find unpaid invoices in QuickBooks or Stripe and draft reminders | Confirm amounts, customer names, and timing |
| Sales follow-up | Draft follow-ups using HubSpot contact and deal history | Confirm deal stage and any promises in the draft |
| Marketing | Turn a product launch into a Canva social campaign | Approve brand claims, visuals, and publishing |
| Payroll | Prepare payroll in Gusto with hours and reimbursements | Review every figure before approval |
| Hiring | Draft a job posting and publish to LinkedIn after edits | Review wording and any compliance requirements |
Anthropic has shared examples from early small business testers. A brokerage owner said Duet flagged a property showing that was on her calendar for the wrong day. A restaurant owner said he used it for accounting, customer support, and marketing, and worked through invoices across multiple locations. A brand strategist said Duet prepared a brief for a client call and noticed the call fell right before her flight. These are company-reported examples, not independent test results.
A realistic end-to-end workflow looks like this: before a client meeting, Duet reviews your calendar, pulls recent emails and relevant CRM notes, and prepares a briefing. After the meeting, you ask it to draft a follow-up email. You read the draft, correct anything inaccurate, and approve it before it is sent. This is a workflow built from Anthropic’s listed capabilities, not an official demonstration.
Claude Duet fits best for individual workflows. Beginners can start with simple administrative tasks. Professionals can use it for meeting preparation and information synthesis. Business owners can use it for reporting and task coordination. Because Team and Enterprise plans are not included yet, teams should not assume it supports organizational deployment.
Claude Duet Pricing, Subscription Requirements, and Usage Limits
Claude Duet has no separate fee during the research preview. It is included in Pro and Max plans for US users. Access still depends on being admitted from the waitlist.
| Plan | Current price (per pricing page) | Claude Duet during the research preview |
| Free | $0 | Not included |
| Pro | 17/monthbilledannually(200 up front, or $20/month billed monthly | Included, US waitlist |
| Max | From $100/month | Included, US waitlist |
| Team | $20 per standard seat/month billed annually | Not included yet |
| Enterprise | $20 per seat/month plus usage at API rates | Not included yet |
Duet shares your plan’s usage limits with the rest of Claude, but it uses them differently from a typical conversation. Background checks count toward your limits, and larger tasks such as building a financial model or a deck require more model turns and background work. Anthropic’s pricing page describes usage limits that reset on a rolling five-hour session window, with weekly limits on paid plans. If you run heavy delegated work, expect your allowance to go faster.
A qualifying subscription does not guarantee admission. Check the official pricing and product pages before subscribing, since Anthropic notes that prices and plans may change.
Is Claude Duet Safe? Privacy, Permissions, and Approvals
Anthropic describes several safeguards for Claude Duet:
- Credential handling: Anthropic says Duet follows the same credential design it uses for Claude Managed Agents, so it does not directly see passwords or sign-in tokens.
- Approval checks: Anthropic says actions are approved by its auto mode classifier, a safeguard it describes as protecting Claude and Claude Code from steps users did not request.
- Prompt injection defenses: Anthropic says Duet’s defenses against hidden instructions in content come from its browser-use research.
- Sensitive information classifier: Anthropic says a dedicated classifier and safety prompt check each message. It is designed to reject requests to share the strictest tier of information, such as credit card numbers and government identification. Health and personal finance information is shared only when you explicitly instruct Duet to send it.
- Testing: Anthropic says red teams and an external firm penetration tested Duet, and it is covered by Anthropic’s public bug bounty program.
These protections reduce risk, but they do not eliminate mistakes or misuse. Anthropic describes them as ongoing work and says it will share more details about Duet’s safety design in the coming days.
Keep your own practices strong. Start with low-risk tasks. Avoid sharing sensitive information you do not need to share. Verify important outputs before acting on them. Review every proposed external action before approving it. If you use Claude Duet for work, follow your organization’s security and data-handling policies.
Claude Duet Limitations: What You Should Know
Anthropic’s official materials describe Claude Duet as an early version. It may misread a thread or misjudge what is urgent. It can miss something that arrives between its checks, so you should not treat it as a real-time monitor. Its personal-errand capabilities are weaker, and booking tickets is an example Anthropic gives. It also cannot yet show you its browser.
Usage is another constraint. Background work draws from your plan’s shared limits, and complex tasks consume more. Team and Enterprise plans are not included yet. Features Anthropic lists as coming soon, including 1Password sign-in and the Shopify storefront editor, are not available now.
Anthropic says the research preview may change based on what early users report. Human review remains necessary for anything that affects money, customers, contracts, or your reputation.
Final Verdict
Claude Duet is most useful for people who spend hours each week on coordination work: triaging messages, preparing for meetings, following up on invoices, and assembling reports. Its strength is that it starts work and reaches out to you, rather than waiting to be asked. Its limits are equally clear. It is a research preview; access is restricted to US Pro and Max subscribers on a waitlist; it uses shared usage limits; and it can misjudge urgency.
If you qualify, check the official product page, confirm your plan and region, and join the waitlist. Start with one low-risk workflow, review everything it produces, and expand from there.
Frequently Asked Questions
What is Claude Duet?
Claude Duet is Anthropic’s proactive AI agent. It sends you a morning brief, takes on tasks you hand it, works in the background, and messages you when work is done or needs your approval.
How do you use Claude Duet?
Install the app or open it on claude.ai after you are admitted, then message it in plain language describing the outcome you want. Review its notifications, check its output, and approve only the actions you have verified.
Is Claude Duet free?
Claude Duet has no separate fee during the research preview. It is included in Pro and Max plans for US users, and it shares those plans’ usage limits.
Who can access Claude Duet?
Anthropic describes access as a waitlist for US Pro and Max subscribers. Joining the waitlist does not guarantee admission.
Is Claude Duet available outside the United States?
Not according to Anthropic’s current official materials, which describe the research preview as starting with US users. Check the product page for updates.
Can Claude Duet work while your computer is off?
Yes. Anthropic says Duet works on its own cloud virtual machine, so your laptop or phone does not need to be on for tasks to continue.
Can Claude Duet send emails or make payments automatically?
No. Anthropic says Duet asks for your approval before sending messages, making payments, signing documents, or deleting anything. Some actions, such as trading securities and moving money between accounts, are off-limits even with approval.
What apps does Claude Duet support?
Anthropic says Duet can use more than 6,000 connectors and plugins in the Claude Marketplace. Its product page lists examples including Gmail, Google Calendar, Slack, Google Drive, QuickBooks, HubSpot, Stripe, and Shopify.
How is Claude Duet different from Claude?
Standard Claude waits for you to open a conversation. Claude Duet comes to you with a morning brief, works in the background on tasks you delegate, and reaches out when something needs you. The two share your account context, but Duet currently lives in its own app.
Is Claude Duet worth trying?
It may be worth trying if you have recurring coordination work across email, calendar, invoices, or reports and you can review its output carefully. If you do not qualify for the preview or you need personal errands handled, the standard Claude experience may serve you better for now.
