TLDRocket
Sign in

Claude Duet

Claude ● Covered by 7 sources

Claude Duet is a new proactive AI agent from Anthropic for Pro and Max users in the US. It works on tasks in the background, then checks in when it needs you.

Based on reporting by Claude — read the original for the full story.

Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error

Anthropic has started a research preview of Claude Duet, a proactive agent built to do work on its own and only bring you in when it has to. It runs on Opus 5.5 and is available to Claude Pro and Max subscribers in the US.

The pitch is simple: hand it a task, then step away. Claude Duet can brief you in the morning, draft next steps, and keep going until it finishes something or needs approval. It can also set up meetings, build dashboards, and return the result in the same thread.

That thread matters. Claude Duet keeps your history, memory, skills, and connected apps together from the first message, so it behaves more like a long-running assistant than a fresh chat every time. Anthropic says it works through text or voice, with no prompts or setup, and can hand back finished work like a doc, sheet, or deck for review.

The company is also leaning hard on safety and control. Claude Duet asks before sensitive actions such as sending messages or paying, and approvals, logins, and safety checks happen in its own cloud computer. Users can see what it has done and steer it along the way.

Anthropic is aiming beyond small errands here. The preview can flag inbox and calendar items, draft replies, find unpaid invoices, prepare payroll for review, and get contracts ready to send. But it is still a preview, and Anthropic says it can misread a thread, get urgency wrong, or miss something that arrives between check-ins. Team and Enterprise plans are out for now, and the waitlist-only rollout keeps the whole thing tightly fenced.

My take — AI-written commentary, not fact-checked reporting

This is the right direction for agent products: fewer circus tricks, more permissioned work that knows when to stop talking. The real test isn’t whether it can make a deck; it’s whether people trust it with the boring, sensitive stuff that actually fills a workday. Most AI demos still want applause. Claude Duet wants access, which is a much tougher sales job.

Read more about this at: Claude

Related stories

The daily briefing

Every AI story that matters, in your inbox by 8am.

TLDRocket reads all relevant sources, removes duplicate coverage, and summarises the day in two minutes. Follow companies and topics for alerts, or get the briefing in Slack. Free, no spam, unsubscribe anytime.