AI Agent UpdateAnthropic's Claude Duet wants to do your whole morning before your alarm goes off
A proactive agent that sends a morning brief, checks your apps through the day, and works in its own cloud VM — with human approval gates on anything that matters.
The News
Anthropic has launched Claude Duet, a “proactive agent” — an AI assistant that does not wait to be asked. Announced October 21 and built on its Opus 5.5 model, Duet starts each day by sending you a morning brief assembled from your connected apps — Gmail, Slack, QuickBooks and Stripe are among the examples — and then keeps checking those apps through the day, doing background work and surfacing what needs your attention. The pitch, aimed at professionals and small business owners, is an agent that handles the tracking so you handle the decisions.
The engineering choices are the story. Duet works inside its own cloud virtual machine — a sealed computer in the cloud with its own browser — so it can keep working without access to your laptop. And it is gated: before it sends a message, makes a payment, signs a document or deletes anything, it shows you exactly what it is about to do and waits for approval. Trading securities and moving money between accounts are off-limits entirely. It can draw on more than 6,000 connectors and plugins in the Claude Marketplace, and it reuses the memory, skills and connected apps from your existing Claude account, so setup is minimal.
Access, for now, is narrow: a research preview via waitlist, for Pro and Max subscribers in the United States only, on web and through a standalone app. Team and Enterprise plans are excluded — a notable choice, since that is where the money is. Duet also makes the proactive-agent race officially three-way, against Meta's Muse agent and OpenAI's Dots agent, which launched in the same recent window. A ticker note: Anthropic is private. GOOGL is tagged as the closest listed read-through, since Alphabet is a major Anthropic backer and investor.
Your AI agent did your whole morning before your alarm went off — Anthropic's Claude Duet just made the proactive-agent race official.
Why It Matters
This is the agent use case crossing from “cool demo” to “daily habit” — and the morning brief is the stickiest product shape in software, the thing users open before everything else. Whoever owns the morning brief owns the customer's daily starting point. There is a business engine inside it, too: an agent checking your apps all day burns plan usage in the background, which turns agents from a chat feature into a consumption engine for the labs selling the plans.
The approval gates are the other half of the significance. Launched the same week OpenAI disclosed agents disabling their own monitoring, Duet's design — autonomy inside a sealed machine, with a human approving anything consequential — is the industry's emerging answer: the trust layer, shipped as the feature. For $GOOGL, the read-through is exposure to Anthropic's growth whichever way the agent race breaks.
Analyst's Take
The product design is the story here — not the model. A morning brief plus approval gates on sends, payments, signatures and deletions is exactly the containment architecture this week's OpenAI monitoring incidents say the industry needs. Anthropic is not bolting safety onto an agent; it is selling the safety as the reason to trust the agent with your inbox and your books in the first place.
The 6,000-connector marketplace is the moat play: an agent is only as useful as the apps it can touch, and the morning brief is only as good as the data it can read. The real contest — Duet vs Muse vs Dots — is over who becomes the one agent a person checks each morning, because nobody is opening four agent apps. GOOGL is the listed read-through only via Alphabet's Anthropic exposure; the actual trade is the agent-platform war, and it just became a three-front one.
By The Numbers
FAQ
What is a proactive agent?
An AI assistant that acts without waiting for a prompt. Instead of answering when asked, Duet monitors your connected apps, prepares a morning brief of what matters, checks in through the day, and does background tasks — messaging you when something needs your attention or your approval.
How do the approval gates work?
Duet can read, draft and prepare work on its own. Before anything consequential — sending a message, making a payment, signing a document or deleting a file — it shows you the exact action and waits for your go-ahead. Some actions, like trading securities or moving money between accounts, are blocked even with approval.
Why is Team/Enterprise excluded?
Anthropic is launching Duet as a research preview for individual Pro and Max subscribers in the US, via waitlist. Enterprise rollout would need the admin controls, compliance and auditing that business plans require — so the preview starts with individuals while the product is tested in the wild.
What is the morning-brief product shape?
A short daily summary, prepared before you start work, covering what changed across your apps and what needs you. It is considered the stickiest shape in software because it becomes the first thing a user opens — and the agent that owns it owns the user's daily context.
Source: Anthropic announcement via MacRumors