Anthropic Enters the Personal-Agent Race: Claude Duet, a Proactive Approval-Gated Agent Built on Opus 5.5

Anthropic Enters the Personal-Agent Race: Claude Duet, a Proactive Approval-Gated Agent Built on Opus 5.5

New subject (announced Wednesday, Oct 21, 2026). Anthropic opened a research preview of Claude Duet, its proactive personal agent — the last of the big labs to enter the category. The Verge: "Anthropic just announced an AI agent competitor to Meta's Muse and OpenAI's Dots: Claude Duet." It is available as a standalone app on iOS and Android and through Claude on the web, built on Claude Opus 5.5, pitched at professionals and small-business owners.

The product shape, per Anthropic: each morning it sends a brief on the most important items across connected apps; it "keeps tabs on those apps and messages you only when something needs your attention"; and it handles tasks "from simple administrative tasks like scheduling meetings to more complex tasks like building dashboards, creating financial forecasts, and developing web apps." Duet inherits memory, skills, and connected apps from Claude, and runs with "a virtual machine with a browser" for completing work. The safety gating is the differentiator: "Before performing significant actions (like sending a payment or deleting data), Claude Duet requests your approval¹," plus a new safety system designed to reject requests to share highly sensitive data, "like credit card numbers."

Anthropic is candid about the limits: Duet is an "early version" that "may misread threads or misassess what is urgent," and since it is workplace-first, personal errands like booking tickets are "not quite as strong" (CNBC). Reuters adds it "requires user approval for sensitive actions, while acknowledging the agent can misjudge urgency."

Why it matters. The proactive personal-agent category is now a four-way race — Meta's Muse (Meta's Muse Wonder (Codename "Watermelon"): Third-Party Evals Confirm a Real Frontier Jump — #3 on Artificial Analysis, Still Behind Anthropic's Top Two), OpenAI's Dots, xAI's Grok Bot, and now Duet — and Anthropic's entry leans explicitly on approval-gating and workplace focus, consistent with the safety positioning it is taking into its IPO. The open question is habit, not capability: the first substantive hands-on review was titled "Claude Duet is beautiful. I still forgot to use it" — a proactive agent that never nudges you is a growth problem.