Anthropic launches proactive AI agent Claude Duet

Anthropic Launches Claude Duet, a Proactive AI Agent

Anthropic has rolled out Claude Duet as a research preview: a standalone app powered by Opus 5.5 that runs on its own cloud computer, sends a morning brief, checks connected apps every few hours, and only pings the user when a decision is needed.

Anthropic unveiled Claude Duet on October 21, 2026: a standalone app where Claude works as a proactive agent instead of waiting for prompts. It is Anthropic’s answer to Meta’s Muse agent and OpenAI’s Dots agent. All three point at the same shift: AI that reaches out before you ask.

Duet is powered by Claude Opus 5.5 and runs on its own cloud computer. It does not need the user’s device to be on; it keeps working in the background. Anthropic says it built Duet for professionals and small business owners with too much to track — the people most likely to find it useful.

The centerpiece is a morning brief. Each morning Duet reviews connected email and calendar accounts, surfaces items that may need attention, and prepares next steps, such as drafting email replies. During the day it checks connected apps every few hours and sends an alert when a task is done or a decision is needed. Users talk to it in one ongoing thread — by text or voice — instead of starting a new conversation for every task.

Duet carries over memory, skills, and connected apps from the user’s existing Claude account, so there is little setup. According to Anthropic, it works with more than 6,000 connectors and plugins in the Claude ecosystem, including Gmail, Google Calendar, Google Drive, Slack, QuickBooks, Stripe, HubSpot, DocuSign, Gusto, Canva, and Shopify. For websites without a connector, it can use its own browser — though users cannot currently watch that browser or take control of it mid-task.

The work it can take on includes research briefs, financial models, sales forecasts, slide decks, dashboards, prototypes, and web apps. Anthropic’s examples: spotting unpaid invoices and preparing reminder emails, filling in payroll information for review, preparing agreements for signature, drafting follow-up emails from CRM history, and helping run a Shopify store.

Guardrails and limits

Duet asks for approval before consequential actions: sending messages, making payments, signing documents, or deleting information. Some actions stay off-limits even with approval — securities trading and moving money between financial accounts are explicitly blocked.

Purchases run through Stripe’s Link digital wallet and Shopify’s Shop Pay, each requiring user approval. Anthropic says Link purchases use a one-time virtual card, so merchants never see the user’s underlying card number.

On access control, Anthropic says Duet does not see passwords or logins when given account access, and a dedicated classifier plus a safety system prompt checks each message to guard against sensitive information leaks.

Availability

Claude Duet is a research preview with access granted gradually through a waitlist. It is limited to Pro and Max subscribers in the United States at no extra cost, but it shares existing plan limits — background checks count toward them. Team and Enterprise plans are not supported at launch. It is available on the web and as a standalone Duet iOS app.

Proactive AI assistants have become the new front of late 2026. Meta launched its Muse agent in late September, OpenAI followed with Dots, and Anthropic has now answered with Duet. All three are betting on the same thing: the AI that gets to the work first wins the user’s trust.