Anthropic entered the intensifying AI-agent contest on Wednesday with Claude Duet, a work-focused assistant it is positioning against Meta’s Muse, OpenAI’s Dots, Google’s Gemini Spark and newer start-up offerings.¹

The company’s wager is that professionals and small-business owners need less help booking dinner and more help managing the sprawl of messages, schedules and connected workplace tools. Duet, built on Claude Opus 5.5, is designed to deliver a morning brief, monitor linked apps during the day and raise issues that need attention.²

Addy Osmani described the app as a “proactive agent” running across “6,000 connectors/plugins,” saying it continues working after a user steps away.Addy Osmani (addyosmani)³ Anthropic says it can tackle tasks from meeting scheduling to dashboards, forecasts, prototypes and web apps—while checking with users before consequential moves such as sending payments or deleting material.²

That supervision is central to Anthropic’s pitch, and to its caveats. Duet uses a dedicated classifier and safety prompt intended to block sharing of highly sensitive data, including credit-card numbers. Yet the company also warns the early system may “misread a thread or miscalibrate what’s urgent.”²

The limits sharpen the contrast with the consumer-oriented race already underway. OpenAI and Meta have framed their agents around personal tasks such as reservations, whereas Anthropic says Duet was built for work and is less capable at personal errands.¹ It is launching only as a research preview with a waitlist; US Pro or Max subscribers are eligible to use it.²

That restraint has not stopped the competitive pressure. Duet’s arrival follows a rush toward always-on agents that can act independently, and it comes as Meta’s Muse had surpassed 14 million installations, according to Apptopia estimates.⁴ Anthropic is offering a different promise: not an all-purpose digital life manager, but an AI colleague that knows when to interrupt—and, just as importantly, when to ask first.⁴