Claude Duet is an agent that messages you first. Watch the usage meter.
SummaryAnthropic opened a waitlist research preview of Claude Duet, a separate app where Claude sends a morning brief, checks your connected apps every few hours and runs tasks on its own cloud computer. It asks before it sends, pays, signs or deletes. The honest caveat: those background checks spend the same usage limits as the rest of your Claude plan.
· MODEL Editors · 2 min read
Some links may be affiliate links. Disclosure

On October 21 Anthropic opened a research preview of Claude Duet, "a new app where Claude works as a proactive agent." The pitch is a reversal of how chat assistants usually work: instead of waiting for you to open a conversation, Duet comes to you with what it thinks needs your attention.
What it does. Each morning Duet sends a short brief on the most important items across your connected apps. During the day it checks those apps and messages you only when something needs you. You can also hand it tasks by text or voice in one ongoing thread, from scheduling meetings to building dashboards, forecasts, prototypes and small web apps. Anthropic says it runs on a virtual machine with its own browser, so your laptop or phone doesn't need to be on while it works. It is built around Claude Opus 5.5, and it carries over the memory, skills and connected apps from your Claude account, so there's little setup. Anthropic says it can use the more than 6,000 connectors and plugins in the Claude Marketplace, with examples like Gmail, Slack, QuickBooks, Stripe, Gusto, Docusign, HubSpot, Canva and Shopify.
The guardrails. Duet reads, drafts and prepares on its own, but before it sends a message, makes a payment, signs a document or deletes anything, Anthropic says it shows you exactly what it will do and waits for your approval. Some things are off-limits even with approval, such as trading securities or moving money between accounts. Anthropic says it never directly sees your passwords or sign-in tokens, and that a new classifier rejects requests to share the strictest tier of information, such as credit card numbers and government IDs. Purchases go through Link, Stripe's wallet, with a virtual card.
Who it's for. Anthropic says it built Duet "for professionals and small business owners with more on their plates than they can track." It's a waitlist preview for Claude Pro and Max subscribers in the US, as a standalone app on iOS and Android and on claude.ai. Team and Enterprise plans aren't included yet.
The honest caveat. Duet isn't free on top of your plan. During the preview it shares your plan's usage limits with the rest of Claude, its background checks count toward them, and big jobs like a financial model or a deck take more turns. Anthropic says Duet checks your accounts every few hours rather than constantly for exactly that reason, which also means it can miss something that arrives between checks. Anthropic also says it "might misread a thread or miscalibrate what's urgent," personal errands are weaker than work tasks, and you can't yet watch its browser.
MODEL has not tested Claude Duet hands-on; we're not on the waitlist yet. Everything above comes from Anthropic's announcement, product page and FAQ, read on October 21. This is not sponsored and has no affiliate links.
Sources
- Introducing Claude Duet (Oct 21, 2026)
Anthropic
- Claude Duet product page and FAQ
Anthropic
- Announcement thread
Claude on X
FREE · EVERY WEEK
The top AI tools this week, tested.
Every week MODEL's editors pick the AI tools worth your time, try them hands-on, and tell you what's good, what's not, and who it's for. Free, in your inbox.
Free. Unsubscribe anytime. Our picks are editorial and can't be bought.
The weekly three is editorial. A subscription does not buy a place in it.
Continue

Builder tip
How to check your Python environment for a malicious .pth file, and dodge the next poisoned release
Today's LiteLLM 1.107.0 stealer ran from a .pth file, which Python executes on every start, even when you only run pip. Here is a one-line check that starts no Python, a harmless demo we ran ourselves, and two install settings that would have kept that release off your machine.
· MODEL Editors

Hot take
Your AI fixed the bug. Trail of Bits counted what else broke.
Trail of Bits reviewed every agent-written security patch its engineers got merged upstream: 21 confirmed regressions across more than 130 pull requests. None broke a security protection. The lesson isn't that agents can't patch. It's that a patch without proof is just a claim.
· MODEL Editors

The test
DuckDB 2.0, tested: faster, and built for your coding agent
DuckDB's biggest release ran our 8 test queries 1.7 times faster than 1.5.6 with the same answers. Its new agent mode gave an AI agent whole results in 30% fewer bytes. The catch: 1.x can't open its files, and dividing an integer by zero is now an error.
· MODEL Editors