By Hamza Faisal
Anthropic is done waiting for you to ask. The AI lab has unveiled Claude Duet, a standalone agent app that watches your connected apps, sends you a morning briefing, and gets work done before you ever type a prompt — its most aggressive move yet into always-on personal agents.
Built on Claude Opus 5.5, Duet is rolling out as a waitlisted research preview on iOS, Android and claude.ai, initially for US subscribers on Pro and Max plans. Anthropic says the app checks your connected services every few hours and surfaces what needs attention — from scheduling meetings and building dashboards to preparing financial forecasts and managing online stores.
The agent that acts before it's asked
The pitch is a reversal of the chatbot era: instead of you opening an app and typing, Duet works in the background and notifies you when something requires action. It keeps working even while your phone or computer is off, since the agent runs on Anthropic's own cloud infrastructure and checks in periodically.
Duet plugs into more than 6,000 connectors in the Claude Marketplace, including Gmail, Slack, LinkedIn and Shopify — so it can scan your inbox, flag threads that need replies, and even make purchases through Stripe's Link digital wallet and Shopify's Shop Pay. Anthropic says sensitive actions like sending messages, making payments, signing documents or deleting files still require explicit user approval.
OpenAI and Meta already moved
Anthropic is not first to the idea. OpenAI launched always-on "dots" agents last month that pursue user goals across apps like Slack and Microsoft Teams, while Meta's Muse AI agent drew millions of downloads and kicked off a fresh wave of investor interest in autonomous AI services. Duet is Anthropic's answer — its consumer-grade play in a race that is shifting from chatbots you prompt to agents that act unprompted.
The launch also shows how quickly the agent economy is maturing: the labs that once competed purely on model benchmarks are now competing on who can quietly do the most of your work for you.
Anthropic admits it's not perfect yet
The company is unusually candid about Duet's limits. Anthropic says the agent can misjudge urgency — flagging things that don't matter or missing things that do — and remains less effective at personal tasks such as booking tickets. It's a useful reminder that an agent acting on behalf of someone is only as reliable as its judgment calls.
Still, with background checks counting toward plan limits and waitlist demand building, Anthropic is clearly betting that proactive is the next default mode for AI assistants — and that users will trust an agent that starts the workday before they do.
By Hamza Faisal
