Last month, Meta’s Muse surged into the personal-agent race, with Mark Zuckerberg saying more than 10 million people in the US and Canada had set up Muse agents in roughly six weeks. Meta has positioned Muse chiefly for consumers, while OpenAI’s Dots spans consumer and enterprise use.¹

Wednesday: Anthropic answered with a different wager: Claude Duet, a research-preview agent built for professionals and small-business owners rather than everyday errands. The company says it can assemble morning briefs from connected apps, flag issues through the day, schedule meetings, build dashboards and produce forecasts. But it concedes personal tasks such as booking tickets are “not quite as strong.”¹

Anthropic’s pitch is not merely that Duet can work unattended, but that it can know when not to. The agent is designed to keep working after a user steps away and to check before acting on their behalf, according to the launch announcement.Addy Osmani (addyosmani)² For consequential steps—including sending payments or deleting material—it asks for approval; Anthropic says it also uses a dedicated classifier and safety prompt to block requests involving its most sensitive information tier.³

That caution is central to the workplace-first argument. A missed reservation is irritating; a badly judged financial action or mishandled company data is more serious. Anthropic nevertheless warns that this is an early version that “might misread a thread or miscalibrate what’s urgent.”³

The product is powered by Opus 5.5 and offered through a waitlist to eligible Pro and Max subscribers, via standalone mobile apps and the web. Its arrival intensifies a crowded market: Meta is chasing mass consumer adoption, OpenAI is straddling both worlds, and Anthropic is betting that the agent users retain will be the one they trust with the workday.⁴