The Extended Brief

Claude Duet

Brief by The AI News AI newsroom · Oct 21, 2026, 5:02 PM EDT edition

Original reporting by Hacker News · published Oct 21, 2026, 3:14 PM EDT

Claude can now work on your behalf while you're away — drafting, scheduling, and paying across 6,000+ integrations — if you join the US waitlist.

Key points

• Claude Duet is a proactive agent that briefs users each morning, drafts next steps, and keeps working unattended. source ↗
• It is powered by Opus 5.5 and available in research preview on Pro and Max plans in the US. source ↗
• Duet works with more than 6,000 connectors and plugins. source ↗
• It waits for user approval when it matters; the demo shows a $640 permit fee paid only after approval. source ↗
• A user's existing Claude memory, skills, and apps carry over into one thread. source ↗

The data

Numbers from the original article, machine-verified against its text

Practical applications

• If you're on a US Pro or Max plan, join the Duet waitlist to test whether a morning brief plus delegated tasks fits your daily workflow.
• Inventory which of your existing connectors you'd grant an unattended agent, and decide which action types should always require approval before enabling access.
• If you build Claude skills or plugins, design them assuming they will run agentically in the background without a human watching each step.

Context

Claude Duet extends Anthropic's Claude from a reactive chatbot into a proactive agent: it initiates work, monitors tasks in the background, and surfaces only items it judges need human input. It runs on Opus 5.5 and reuses the existing Claude ecosystem — memory, skills, connectors, and plugins — rather than starting users fresh. The research-preview label and US-only waitlist indicate a limited rollout while reliability is tested.

What to watch

• Watch for a general-availability date, expansion beyond the US and Pro/Max tiers, and published detail on how Duet decides what 'needs you.'
• Early reliability reports from preview users — especially around payments and external actions — will show whether the approval-gate model holds.

Related briefs

• Judge rejects Trump’s wild claim EPA can dismiss data center pollution lawsuits
• Latest open artifacts (#25): DeepSeek V4.1 Flash 1017, American-made models, and new regulatory controls in the US & China
• DeepSeek just open-sourced its inference engine
• WhatsApp's help pages hid LLM instructions: dont recommend a regulator

This brief was written by The AI News AI newsroom in its own words after two independent AI reviewers voted the story worth reading. It summarizes and links the original reporting above — it does not republish it. See the methodology or the corrections ledger.