Anthropic announced on October 21, 2026 the opening of a research preview of Claude Duet, a new app where Claude acts as a proactive agent. The system comes to the user with what needs their attention, completes requested tasks, keeps working even when the person steps away and asks for approval before acting on their behalf.
Claude Duet is available as a standalone app for iOS and Android, plus access via claude.ai. In this preview phase, it is included in Pro and Max plans and uses the same usage limits as the rest of Claude. Initial access is restricted to subscribers in the United States and requires joining a waitlist.
What happened?
The company positions the product for professionals and small business owners who manage many decisions throughout the day. The agent was built around the Claude Opus 5.5 model and carries over memory, skills and connected apps from the existing Claude account.
Every morning it sends a short briefing with the most important items from connected apps. Throughout the day it monitors those accounts every few hours and only notifies when something requires attention. Tasks range from scheduling meetings to building dashboards, financial forecasts or web apps. It runs on its own virtual machine with a browser, so it does not require the user's device to be on.
The agent connects to more than 6,000 connectors and plugins in the Claude Marketplace. Cited uses include: finding missed messages in Gmail and Slack and helping reply, checking invoices in QuickBooks or Stripe, preparing payroll in Gusto, drafting job posts on LinkedIn, preparing documents in DocuSign, creating campaigns in Canva and managing Shopify stores. Online purchases can be made via Stripe Link or Shop Pay, always with approval.
Why it matters
The launch arrives amid a race for AI agents. Anthropic positions itself against offerings such as Meta Muse and OpenAI Dots, emphasizing focus on professional work rather than personal tasks. The system requires explicit approval to send messages, make payments, sign documents or delete anything. It does not perform operations such as trading securities or transferring money between accounts.
There is a dedicated classifier that rejects requests to share sensitive information such as credit card numbers or government identification. The company acknowledges that the early version may still misjudge the urgency of a situation and that personal tasks, such as booking tickets, are not its strength. Background checks count against the plan's limits.
Context
The official announcement was made by Anthropic on October 21, with coverage in outlets such as TechCrunch and Reuters. The company highlights years of work on agent safety, including adversarial tests and a bug bounty program. What is learned from early users should influence future integration with the main Claude app.
Sources
- Official Anthropic announcement: Introducing Claude Duet
- Claude Duet product page
- TechCrunch: Claude Duet app is Anthropic's bet on agents
- Reuters: Anthropic unveils Claude Duet AI agent app
Transparency: This content was created, edited or reviewed with the assistance of artificial intelligence. The information was cross-checked with public posts on X and sources available on the internet. Consult the original sources to verify the full context.
By GeekikiBot
