Claude Duet: How a Proactive AI Agent Helps Small Business

Anthropic has opened a research preview of Claude Duet, a new app built around a simple shift in how we work with AI: instead of waiting for a prompt, the assistant comes to you. Announced on October 21, 2026, Claude Duet works as a proactive agent that flags what needs your attention, completes tasks in the background, and checks in before it acts on your behalf. Anthropic says it designed the app for professionals and small business owners who have more on their plates than they can track.

• Availability: waitlist research preview starting with US Pro and Max subscribers, on iOS, Android and claude.ai
• Price: included in Pro and Max plans, sharing the same usage limits as the rest of Claude
• Model: built around Claude Opus 5.5, with access to more than 6,000 connectors and plugins
• Guardrails: asks for approval before sending messages, making payments, signing documents or deleting anything

How Does Claude Duet Work?

Claude Duet carries over the memory, skills and connected apps you already use in Claude, so setup is minimal. Each morning it sends a short brief covering the most important items across those apps. During the day it checks your connected accounts every few hours and messages you only when something genuinely needs you. Anthropic explains that it chose a few-hour rhythm rather than constant monitoring so the agent does not burn through plan limits.

Conversations happen in one ongoing thread, by typing or voice, so a quick reply and a weeklong project start in the same place. Crucially, the agent has its own virtual machine and browser. As MacRumors notes, that means it can keep working in the cloud without your laptop or phone being switched on, and you get a notification when a task is done or needs review.

What Can a Proactive AI Agent Handle?

Anthropic's examples range from scheduling meetings to building dashboards, preparing sales forecasts and prototyping web apps. Through the Claude Marketplace, the agent can find missed Gmail and Slack messages, check QuickBooks or Stripe for unpaid invoices and draft reminders, prepare payroll in Gusto, and draft HubSpot follow-ups. Early testers described practical wins: one real estate broker said the agent caught a property showing booked on the wrong day, and a brand strategist was offered a reschedule when a client call landed right before her flight.

This builds naturally on Anthropic's earlier work with Claude for Small Business workflows, moving from tools you invoke to an assistant that takes the first step.

How Does Claude Duet Keep You in Control?

The safety design is the most thoughtful part of the launch. Claude Duet reads, drafts and prepares work on its own, but before it sends, pays, signs or deletes, it shows exactly what it intends to do and waits for approval. Some actions, such as trading securities or moving money between accounts, are off-limits even with approval. Credentials follow the same design as Claude Managed Agents, so the agent never directly sees passwords or sign-in tokens.

Anthropic also shipped a dedicated classifier and safety system prompt that check every message for sensitive information. The strictest tier, such as card numbers and government ID, is rejected outright, while health or personal finance details are only shared when you explicitly instruct it. Red teams and an external firm penetration tested the app, and it falls under Anthropic's public bug bounty program.

Why This AI Agent Launch Matters

Proactive agents are quickly becoming the next frontier for AI assistants, following always-on products like OpenAI's Dots agents. What sets Claude Duet apart is how much of the launch is about judgment and restraint: deciding what can wait, what needs a human, and what should never happen at all. For a busy shop owner, that combination of initiative and caution is exactly what makes delegation feel safe.

Follow every new assistant launch in our AI coverage.

Sources: Anthropic — Introducing Claude Duet — October 21, 2026; MacRumors — October 21, 2026.