Anthropic Launches Claude Duet AI Agent to Automate Business Tasks

Anthropic has launched Claude Duet, a new artificial intelligence application designed to help professionals and small-business owners manage their daily responsibilities with less manual effort. Announced on October 21, 2026, the product moves beyond the traditional chatbot model by allowing Claude to identify important tasks, prepare daily briefings and continue working even when users are away from their devices. Built around Anthropic’s Claude Opus 5.5 model, Duet can assist with activities ranging from scheduling meetings and organising communications to developing financial forecasts, dashboards and web applications. The company says the goal is to create an AI assistant that does not simply wait for instructions but can recognise when something needs attention and help users respond.

One of Duet’s main features is its ability to connect with other applications and gather information from different parts of a user’s working environment. Anthropic says the agent can access more than 6,000 connectors and plugins available through the Claude Marketplace, including tools used for email, customer management, accounting and online commerce. It can review Gmail and Slack messages, identify unpaid invoices through QuickBooks or Stripe, prepare customer follow-ups using HubSpot information and help organise recruitment activities. It can also assist with creating marketing campaigns in Canva and managing Shopify stores. By bringing information together, Duet aims to reduce the time people spend switching between applications and manually coordinating routine business activities.

Unlike conventional AI assistants that operate only during active conversations, Claude Duet can continue working through its own virtual machine and browser. The application can prepare a morning briefing using connected services and periodically check for developments that may require the user’s attention. Users can also assign more complex tasks, such as preparing a financial forecast or building a dashboard, and receive a notification when the work is completed or requires a decision. However, Anthropic says the system checks connected accounts every few hours rather than continuously monitoring every activity in real time. The company also acknowledges that the early version can misunderstand conversations or misjudge the urgency of particular issues, meaning users should review important recommendations rather than rely on the agent without supervision.

Security and user approval are central to the product’s design because Duet can interact with sensitive information and external services. Anthropic says the agent must request approval before performing consequential actions such as sending messages, making payments, signing documents or deleting information. Certain activities, including securities trading and transferring money between financial accounts, remain prohibited even with user approval. The company has also introduced additional safeguards intended to prevent sensitive information, such as payment-card numbers and government identification details, from being disclosed improperly. According to Anthropic, its security measures include protection against malicious instructions hidden in online content, credential isolation and testing by internal security teams and an external security firm. Nevertheless, these safeguards cannot guarantee that every action will be correct or that every security risk will be eliminated.

The launch places Anthropic in direct competition with other companies developing AI agents that can independently carry out tasks. OpenAI has introduced Dots, while Meta is expanding its Muse assistant, with both companies pursuing systems that can work across applications and continue operating in the background. Claude Duet is initially available as a research preview through a waitlist for eligible Pro and Max subscribers in the United States, with access offered through standalone iOS and Android applications and Claude on the web. The service shares the usage limits of a user’s existing Claude subscription, meaning background checks and larger assignments consume part of the same allowance. Anthropic has not announced a general release date for all users. The success of Duet will ultimately depend on whether it can reliably complete useful work, protect sensitive information and save enough time to justify the additional AI activity.