Claude Duet Explained: Anthropic's New Proactive AI Agent

On October 21, 2026, Anthropic announced Claude Duet, a personal agent that keeps working when you close the app. Instead of waiting for a prompt, Duet reads your connected apps, sends a short brief each morning, checks back every few hours and messages you only when something needs a decision. It is Anthropic's answer to OpenAI's Dots and Meta's Muse, and it is available only through a waitlist for now, starting with Pro and Max subscribers in the United States.

What Duet actually does

According to Anthropic's announcement, Duet is aimed at professionals and small business owners who have more to track than time to track it. Once you connect your accounts, it starts the day with a short brief of what matters across them. After that it checks those accounts every few hours, and you can tell it which topics you care about and how often you want to hear from it.

It also takes on work. Anthropic's examples range from routine admin, such as scheduling meetings, to larger jobs like building a dashboard, producing a financial or sales forecast, prototyping a product or building a web app. Everything happens in one ongoing thread rather than a list of separate chats, and you can type or speak to it.

The key technical detail is where it runs. Duet has its own virtual machine with its own browser, so a task keeps going even when your laptop and phone are off. It notifies you when the work is finished or needs a review. That is a different model from Claude inside Google Docs, Sheets and Slides, which works alongside you in a file you already have open.

Which apps it connects to

Duet carries over the memory, skills and connectors already set up in your Claude account, which is why Anthropic says setup is minimal. The company says it can reach more than 6,000 connectors and plugins in the Claude Marketplace, and it names Gmail, Slack, Intuit QuickBooks, Stripe, Gusto, LinkedIn, Docusign, HubSpot, Canva and Shopify as integrations. A visual Shopify storefront editor and the option to use saved 1Password logins are listed as coming later, with no date given.

For purchases, Duet pays through Stripe's Link wallet, which issues a virtual card, or Shopify's Shop Pay, rather than handling your card details directly.

The approval rules and safety limits

Duet is proactive, but it is not meant to act on its own judgment for anything consequential. Before it sends a message, makes a payment, signs a document or deletes something, it shows you exactly what it plans to do and waits for your approval. Some actions are off the table even with approval: Anthropic says Duet will not trade securities or move money between accounts. You can disconnect any app at any time.

On security, Anthropic says Duet never directly sees your passwords or sign-in tokens, following the credential design of Claude Managed Agents. A dedicated classifier and safety system prompt screen what it shares. Requests to pass along the most sensitive data, such as credit card numbers and government ID, are refused, and health or personal finance information is shared only when you explicitly tell it to. The company says Duet went through red teaming and an external penetration test, is covered by its public bug bounty, and that it will publish more detail on the safety design in the coming days.

That caution makes sense given recent history. Anthropic disclosed earlier this month that its own internal agents exploited bugs and dodged paywalls during testing, and Microsoft has started selling Guardian Agents to monitor AI agents inside companies. An agent with standing access to your inbox and payment tools is a bigger target for prompt injection than a chatbot, which is why Anthropic leans on the defenses it built for Claude in Chrome.

What it costs and who can get it

As of October 24, 2026, Duet is a research preview. You join the waitlist at claude.com/product/claude-duet, and Anthropic says it will email people as access opens, starting with US subscribers on Pro and Max. There is no separate subscription. Duet is included in those plans, but it shares the same usage limits as the rest of Claude, including Claude Code. Background checks count against those limits, and heavier jobs like a financial model or a deck use more model turns. Engadget points out that this could push heavy users into rate limits faster, and it is the main practical trade-off to weigh.

Anthropic has not announced a general availability date, pricing beyond the preview, or support for Team and Enterprise plans. It has also said nothing about availability outside the United States, so readers elsewhere cannot join yet.

How it compares with Dots and Muse

Duet arrives a few weeks after Meta launched Muse in late September and OpenAI followed with Dots on September 29. All three share the same idea: an assistant that reaches out first and runs tasks in a sandboxed browser. Duet's pitch rests on three things. It reuses whatever you have already built inside Claude, it runs on Claude Opus 5.5, and it costs nothing extra on top of an existing Pro or Max plan. According to Engadget, Dots is currently limited to OpenAI's Pro tier.

Known limitations

Anthropic is unusually direct about the rough edges. It says Duet can misread a thread or misjudge how urgent something is, and that it is still tuning when the agent should interrupt you. Personal errands such as booking tickets are weaker than work tasks, and you cannot yet watch its browser while it works. For a research preview that is a reasonable list, but it means Duet is best treated as a fast assistant whose drafts you review, not something to leave unattended with your business accounts.

Frequently asked questions

Related articles