Claude Duet: Anthropic's Proactive AI Agent That Messages You First

Introduction: The Agent That Comes to You

For three years, using an AI assistant has meant the same ritual: you open a chat window, type a prompt, and wait. On October 21, 2026, Anthropic flipped that around. Claude Duet, opened as a research preview this week, is an AI agent designed to message you — surfacing what needs attention, carrying out assignments in the background, and checking in before it takes any action with real-world consequences.

The launch makes Anthropic the third major US AI lab to ship a consumer-facing agent built to run unattended, after Meta's Muse and OpenAI's Dots. For anyone choosing AI tools right now, Duet matters because it previews where the entire assistant category is heading: from chat windows you open to proactive colleagues that open the conversation. Explore our full directory of AI agent tools to see the wider landscape.

What Is Claude Duet?

Claude Duet is a standalone app — iOS, Android, and claude.ai — where Claude acts as a proactive agent rather than a chat window you open. Anthropic sums up the difference in four verbs: Duet brings you what needs your attention, completes tasks you hand it, keeps working when you step away, and checks with you before it acts on your behalf.

It's built around Claude Opus 5.5, the model Anthropic says gives Duet the judgment to decide what needs you now, what can wait, and when to check in first. The target user is professionals and small business owners with more on their plates than they can track: think unpaid invoices, unanswered customer emails, and tomorrow's scheduling conflict — spotted and pre-handled before you ask.

A key convenience: Duet carries over the memory, skills, and connected apps from your existing Claude account, so there's very little to set up if you already use Claude.

How Duet Works: Briefings, Background VMs, and Approvals

A morning brief, then quiet monitoring

Each morning Duet sends a short rundown of the most important items across your connected apps. Throughout the day it checks those accounts every few hours and messages you only when something actually warrants attention. You control which subjects it flags and how often it updates — the periodic cadence exists partly because constant checking would burn through your plan's usage limits much faster.

Its own cloud computer

Underneath the chat, Duet runs on its own cloud virtual machine with a browser. You describe a task by text or voice in one persistent thread — a quick reply and a weeklong project start in the same place — and Duet works out which applications and steps the job requires, performs the work remotely, and notifies you when it's done or needs a decision. Because the compute lives in the cloud, tasks keep running after you close your laptop or switch off your phone.

Approval gates before anything consequential

Duet asks permission before it sends a message, pays for something, signs, or deletes. Anthropic's examples include drafting a recruitment listing and publishing it to LinkedIn only after you review and authorize it, and making purchases through Stripe's Link wallet or Shop Pay with a virtual card, so your real card number is never shared. You can disconnect any connected app at any time.

With more than 6,000 connectors and plugins available through the Claude Marketplace, the permissions attached to each connection become consequential: access sufficient to summarize an invoice is very different from authority to initiate a payment.

How Duet Differs from Cowork and Remote Control

Duet isn't Anthropic's only agent product, and the distinctions matter. Remote Control gives your phone access to a local coding agent running on your own machine — which must stay awake. Cloud Cowork executes requested or scheduled assignments. Duet's differentiator is upstream of both: it aims to notice work that deserves attention in the first place, rather than waiting for you to request or schedule it. Anthropic is also merging chat and Cowork into a single Claude experience, with Duet as the proactive layer on top.

Claude Duet vs Meta Muse vs OpenAI Dots

All three products share a pattern: a background agent on cloud infrastructure, connectors into your tools, and human approval gates before consequential actions. The differences are in reach and price.

Duet's $20 monthly entry point undercuts the $100 minimum for OpenAI's Dots, while Muse offers a free entry. OpenAI's Dots currently has the clearer organizational access route — something governance-minded buyers should weigh.

Pricing and Access: Who Can Use It Today

One trade-off deserves attention: Duet activity counts against the same usage allowance as your regular Claude conversations. Anthropic hasn't published how much a typical background task consumes, so heavy Duet use could eat into the quota you'd otherwise spend chatting.

What Duet Still Can't Do

Anthropic is candid that this is an early version. The core is built around work, so personal errands like booking tickets aren't as strong. Duet also can't yet show you its browser, which limits visibility into that portion of its work — you see the results, not the steps. And reliability at scale is unproven: no benchmark results, task-level cost estimates, or quantified productivity claims accompanied the announcement.

What This Means for AI Tool Users

Three practical takeaways if you're picking tools this quarter:

• The chat window is becoming the fallback, not the interface. Expect every major assistant to add proactive monitoring and briefings within months.
• Permissions are the new security perimeter. An agent that can summarize your inbox is low-risk; the same agent with payment authority is a different threat model. Grant the least access that gets the job done.
• Usage allowances are the hidden cost. Background agents consume your quota invisibly. Watch your plan limits during any research preview.

Whether Duet, Muse, or Dots wins, the direction is settled: your next AI assistant won't wait for you to open it.

Frequently Asked Questions