Claude Duet Arrives as a Research Preview, Not a Finished Agent

Anthropic has released Claude Duet, an AI agent built on its Claude Opus 5.5 model, as a research preview rather than a finished product. That label matters for anyone deciding whether to adopt it now: Anthropic itself is framing Duet as an early, still-evolving tool, not a stable release aimed at mass deployment.

The agent targets professionals and small business owners who want ongoing oversight of their digital workload instead of a one-off chatbot session. According to Anthropic's own description, reported by Jay Peters at The Verge, Duet sends a short morning brief covering the most important items across a user's connected apps, then monitors those apps throughout the day and messages the user only when something actually needs attention.

Access is broader than many research previews get. Duet is available as a standalone app on iOS and Android, and also through Claude on the web, so users are not restricted to a single platform while Anthropic gathers feedback. That multi-surface rollout suggests Anthropic wants real usage data from varied contexts, phone notifications during a workday, browser tabs at a desk, rather than a narrow lab test.

The competitive framing is explicit. The Verge positions Duet as Anthropic's answer to Meta's Muse and OpenAI's Dots, two other agent products aimed at the same niche: tools that sit across a person's apps and surface what matters without requiring constant prompting. All three companies are effectively racing to define what a persistent, semi-autonomous work assistant looks like before any of them has a long track record of reliability at that job.

What remains untested in public is how Duet behaves when the task list gets messy. Anthropic describes handing it tasks directly, beyond the passive morning-brief function, but a research preview by definition has not been stress-tested across the full range of connected apps, permissions, and edge cases that real small businesses will throw at it. The gap between a tidy product description and a dependable daily tool is exactly what the research-preview label is meant to signal, and it is the gap Anthropic has not yet closed.