Anthropic Launches Claude Duet Proactive AI Agent
Original: Claude Duet
Why This Matters
Proactive agents with approval gates mark a shift from reactive chatbots to autonomous AI workflows.
Anthropic has launched Claude Duet, a proactive AI agent in research preview for Pro and Max plan users in the US. Powered by Opus 5.5, it integrates with 6,000+ connectors, handles tasks autonomously, and briefs users each morning.
Anthropic has introduced Claude Duet, a proactive AI agent currently in research preview, available to Pro and Max plan subscribers in the United States via a waitlist. The product is powered by Opus 5.5 and connects to more than 6,000 plugins and connectors.
Unlike on-demand assistants, Duet is designed to act without being prompted. It delivers a morning brief summarizing pending decisions—contract redlines, expiring permits, fundraising data room gaps—then works in the background until it hits something requiring user approval. The approval model is explicit: Duet pauses and waits for a go-ahead on anything consequential, such as a $640 health permit payment or sending a fundraising deck to an investor.
Duet carries over existing Claude users' memory, skills, and app connections from the standard Claude app, meaning no separate onboarding for established users. The product page demos show it handling multi-step workflows: rebuilding a pitch deck with Q3 data, filling county permit forms, pulling data room documents, and drafting refund responses for 27 incorrect delivery orders—all surfaced in a single thread.
The agent is positioned as an always-on chief of staff that keeps working while users are away, checking back only when judgment is needed. No pricing beyond existing Pro and Max plan tiers has been announced.