Claude Duet and Email: What Anthropic's Agent Does With Your Inbox (2026)

The short version: Claude Duet is Anthropic's new app where Claude works as a proactive agent. It opened as a research preview on October 21, 2026.

• Who can get it: Claude Pro and Max subscribers in the US, through a waitlist. Team and Enterprise plans aren't included yet.
• What it costs: nothing extra during the preview. It is included in Pro ($20 a month) and Max (from $100 a month), and it uses your plan's usage limits.
• What it does with email: every few hours it checks the accounts you connect. It sends a morning brief, drafts replies for you to review and follows up on threads that have gone quiet.
• What waits for you: before it sends a message, deletes anything, pays or signs, it shows you exactly what it's about to do and waits for your approval.
• What it isn't: a way to run a team inbox. It works from your own connected accounts.

This guide covers what Anthropic has published, a safe way to start Duet on your email, and what it means for a team that answers customers from a shared Gmail address. Everything here was checked against Anthropic's announcement and help center on October 22, 2026. It is a research preview, so we will update this page as it changes.

Anthropic's announcement, October 21, 2026. Image: Anthropic.

What is Claude Duet?

Claude Duet is a separate app, on iOS, Android and the web at claude.ai, where Claude comes to you with what needs your attention instead of waiting for a prompt. Anthropic built it "for professionals and small business owners with more on their plates than they can track."

What is new compared with chatting with Claude:

• It comes to you first. Each morning it sends a short brief on the most important items across your connected apps, and through the day it messages you only when something needs you.
• It works while you're away. Duet has its own virtual machine with its own browser in the cloud, so your laptop or phone doesn't need to be on.
• One thread, typed or spoken. You message it like a person. You can also open separate channels for different topics, and they share the same memory.
• It brings your Claude setup with it. Duet runs on your Claude account, so the memory, skills and connectors you've set up in Claude are there from the first message.
• The model. It runs on Claude Opus 5.5, and you can't choose a different model.

Here is Anthropic's own introduction:

Who can get Claude Duet, and what it costs

From Anthropic's help center and pricing page, as of October 22, 2026:

The usage point is worth planning for. Anthropic says Duet checks your accounts at set intervals partly so that background work doesn't use up your limits, and bigger jobs, like a financial model or a deck, take more of them.

How to join the waitlist

1. Check you are eligible: a Claude Pro or Max plan and a Claude account in the US.
2. Join the waitlist on the Claude Duet page, with the email address on your Claude account and your company name.
3. Choose your plan and phone, iPhone or Android, and submit. You can also say what you'd hand off first.
4. Open your invitation on your phone when it's your turn, using the Apple or Google account linked to the same email address. Access is tied to that address.
5. Sign in with your Claude account. Your connectors come with you.

What Claude Duet does with your email

Email is where Duet starts each day. Here is what Anthropic says it does, and what it holds back.

By itself:

• It checks your connected accounts every few hours. Something urgent can arrive between checks.
• It sends a morning brief. "Each morning, it looks across your email and calendar, picks what most needs your attention, and can start the first steps, like drafting replies for you to review."
• It drafts replies for you to review, and follows up on threads that have gone quiet.
• It shows its sources. When it gives you a number or a summary, it tells you which email, file or spreadsheet it came from.

Only with your approval:

• Sending email. "With your approval, it can send emails from the accounts you've connected."
• Deleting anything, paying for something or signing a document. If you can't approve straight away, it finishes the rest of the task and brings the approval back to you later.

Not at all:

• Texts or phone calls to other people.
• Accessing someone else's account, trading securities or moving money between accounts, even with your approval.
• Outlook, for now. Anthropic's help center says Microsoft 365 isn't available yet, despite a Microsoft logo among the apps on the product page.

A morning brief, from Anthropic's product page demo. The last item is the kind of customer problem a support inbox would hold. Image: Anthropic.

How Claude Duet's approvals work, and how they differ from Claude's Gmail connector

Duet is built to ask before every consequential action, with no setting described that lets it send without you. Before it sends a message, makes a payment, signs a document or deletes anything, Anthropic says it "shows you exactly what it's about to do and waits for your approval."

That matches Claude's default elsewhere, with one difference. In ordinary Claude chats, the Gmail connector also asks before each send, reply or forward, but on Team and Enterprise plans an owner can let members allow those actions to run without asking each time (Claude's Google Workspace connector help). Duet is Pro and Max only, and Anthropic describes no equivalent switch.

Two more protections matter for email:

• Sensitive information. A new classifier and safety system prompt check each message. They are designed to refuse to share the most sensitive information, such as credit card numbers and government ID. Health or personal finance details go out only when you've explicitly told Duet to send them.
• Prompt injection. Anthropic is candid that email can contain "instructions meant to mislead it." Duet is built to treat what it reads as information, "but no defense is perfect, which is one reason consequential actions need your approval."

An approval card in Anthropic's demo. Email sends get the same treatment. Image: Anthropic.

A safe way to start Claude Duet on your email

Start Duet on reading and drafting, and keep every send as a deliberate approval. This is our recommendation, not Anthropic's, and it follows the same permission ladder we use for any email agent: triage, draft, act with approval, act alone. Anthropic itself warns that Duet "might misread a thread or miscalibrate what's urgent," so read each draft before you approve it.

1. Connect only the email account it needs. Duet uses the connectors on your Claude account, so review what's already connected before you start.
2. Tell it what you want to hear about. Anthropic says you can tell it what to flag and how often.
3. Approve sends one at a time for the first few weeks, and read each draft rather than skimming it.
4. Treat every email as information, not instructions. If a message asks Duet to forward something, share a file or contact someone new, decline the approval.

Three requests to start with (our examples):

Every weekday morning, include in my brief any email from a customer that's been waiting more than a day for a reply. For each one, give the sender, a one-line summary and a draft reply for me to review. Don't send anything.

Find email threads from the last two weeks where I promised to send something ("I'll send", "I'll get back to you") and haven't. List them with the date I promised and draft a short follow-up for each. Don't send anything.

Check Stripe for unpaid invoices and draft reminder emails for me to approve. Keep each one under 80 words and include the invoice number and amount.

The last one is close to an example from Anthropic's own help center.

Can Claude Duet run a team inbox like support@?

Not on anything Anthropic has published. Duet is a personal agent:

• The plans: Pro and Max only. "It isn't available on Team or Enterprise plans yet."
• The accounts: it works from the accounts you connect, and Claude's Gmail connector can only access the Gmail data for the Google account you've connected.
• The off-limits list: it won't access someone else's account, even with your approval.

A team support address needs things a personal agent doesn't provide on its own: one owner per conversation, a warning when two people reply to the same customer, response-time targets, and hand-offs to a named teammate rather than back to whoever owns the agent. Google's own Gmail Help shows how hard this is even for Gemini: delegates of a shared Gmail mailbox can't use Gemini in Gmail at all.

The practical route is to keep the shared inbox in a tool built for teams and let Claude work it through a connector. Drag turns a Gmail address like support@ into a shared board your team works together, and on the Pro plan, Drag's Claude connector lets Claude triage, assign and draft on those boards, asking before it acts. Every change lands back in Drag, where the whole team can see it.

Will Drag's connector work inside Duet? Anthropic says Duet "can use the apps you've already connected in Claude," and Drag's connector is a custom connector you add to your Claude account itself. So we expect it to be there in Duet. We haven't been able to test that yet: Duet is US-only, and we're in the UK. We'll update this page once we have.

Claude Duet vs OpenAI dots vs Gemini agent vs Meta Muse

Duet is the latest of four big personal or work agents launched in six weeks, and each handles email differently:

In the UK and EU, dots are available today through ChatGPT Business Premium or the Enterprise beta, Duet and Muse aren't, and Google hasn't stated regions for the Gemini agent.

Our full comparison of Claude Duet, OpenAI dots, Meta Muse, the Gemini agent and Grok Bot covers price, approvals and team inboxes agent by agent. For the others in depth, see our guides to OpenAI dots and email and the Gemini agent.

Frequently asked questions

Is Claude Duet free?

Not on a free plan. During the research preview, Claude Duet is included with Claude Pro and Max at no extra cost, so a Pro subscriber pays nothing beyond the $20 a month plan. It does use your plan's usage limits, including the five-hour session and weekly limits, and its background checks and morning brief count toward them.

When will Claude Duet be available in the UK and Europe?

Anthropic hasn't given a date. As of October 22, 2026, the research preview is US only, for Pro and Max subscribers on the waitlist, and Anthropic says invitations go out in waves. We'll update this page when Anthropic announces other countries or the Team and Enterprise rollout it says is planned.

Can Claude Duet send emails without asking?

No. Every send waits for your approval, and it sends only from email accounts you've connected. Unlike Claude's Gmail connector on Team and Enterprise plans, where an owner can let members skip the per-send approval, Anthropic describes no setting in Duet that lets it send on its own. If you can't approve straight away, it finishes the rest of the task first.

Does Claude Duet work with Outlook?

Not yet. Anthropic's help center says Microsoft 365 isn't available in Claude Duet during the research preview, and the email connector it names is Gmail, alongside Google Calendar. Some other connectors, such as Slack, work with limits in Duet.

Related guides

• Claude Duet vs OpenAI dots vs Meta Muse vs Gemini agent: price, email and team inboxes, agent by agent.
• How to connect Gmail to Claude: the connector Duet uses for your own mailbox.
• Claude prompts for emails: prompts that work in chat and in Duet.
• How to connect a shared inbox to Claude: give Claude a team inbox through MCP.
• What is agentic email?: the four permission levels and the real risks.

Want Claude on your team's inbox, not only yours? Start a 7-day free trial of Drag. Varley is included on Plus and Pro, and no card is needed.