Anthropic launches Claude Duet, a proactive AI agent app

The latest:

A standalone AI agent that works without waiting for prompts is Anthropic’s answer to the industry’s agent race. The company on Thursday unveiled Claude Duet, an app that delivers a daily briefing and handles tasks from scheduling meetings to building financial forecasts and web apps, Reuters reported. It arrives as OpenAI and Meta push their own autonomous assistants.

Details:

• What it does: Anthropic said Duet monitors a user’s connected apps and surfaces items requiring attention rather than waiting to be asked. Its task range spans meeting scheduling, financial forecasting and building web apps, alongside a daily briefing delivered to the user.
• Where to get it: The app is available on iOS, Android and claude.ai, but only through a waitlisted research preview, according to Reuters. Anthropic did not say when the preview would open more broadly or move to general availability.
• The model: Duet is built around Anthropic’s Opus 5.5 model and is initially available to US subscribers on the Pro and Max plans. The company said the agent shares those plans’ existing usage limits rather than carrying a separate allowance or price.
• The integrations: Duet connects to more than 6,000 connectors in the Claude Marketplace, including Gmail and Slack, Reuters reported. That footprint determines how much of a user’s working day the agent can see and act inside without manual handoffs.
• Payments: The agent can make purchases through Stripe’s Link digital wallet and Shopify’s Shop Pay, giving it the ability to complete transactions rather than stopping at recommendations. Anthropic did not disclose spending caps attached to those purchases.
• The guardrails: Anthropic said Claude Duet requires user approval before taking sensitive actions. The company also acknowledged the agent can misjudge urgency and remains less effective at personal tasks such as booking tickets, an unusually direct admission of limits at launch.
• The competition: OpenAI last month launched always-on dots agents that pursue user goals across apps including Slack and Microsoft Teams, according to Reuters. Meta’s Muse AI agent drew millions of downloads and fueled investor interest in autonomous AI services.
• The market stake: All three products target the same prize: software that completes work autonomously on a user’s behalf rather than answering questions. That shifts the competition from model benchmarks to how reliably an agent can act inside a customer’s existing tools.

Between the lines:

The design choices point at where Anthropic thinks the risk sits. Pairing payment rails like Link and Shop Pay with a mandatory approval step for sensitive actions concedes that an agent transacting unsupervised is the failure mode users fear most. Flagging misjudged urgency and weak personal-task performance at launch sets expectations before the research preview widens.

What’s next

Watch when Anthropic opens the waitlist beyond US Pro and Max subscribers, whether pricing separates from existing usage limits, and how download and retention figures compare with Meta’s Muse and OpenAI’s dots agents.