Claude Duet: Anthropic's Proactive AI Agent (2026)

Updated: October 2026 · 6 min read · By the CreatorSprout team

Image: Anthropic (Wikimedia Commons, public domain)

Launch coverage based on Anthropic's announcement and press reporting (Reuters, MacRumors) — we have not tested this product hands-on. Duet is in a waitlisted research preview at launch, so real-world behavior is still unverified. Facts checked October 2026; confirm on official pages before you buy. No affiliate links in this article.

The headline for creators and small business owners: an AI that watches your inbox, calendar, and tools for you — and asks permission before doing anything sensitive. Here's what the launch actually contains.

What Claude Duet is

Claude Duet is a separate product from the Claude chatbot — a dedicated agent app for iOS and Android plus access on the web at claude.ai. Anthropic built it for professionals and small business owners with too much to track. You assign tasks through text or voice messages — anything from scheduling meetings and drafting email replies to building dashboards, financial forecasts, and web apps.

The key difference from the Claude app you already know: Duet is designed to identify work that needs attention without waiting for a prompt. Interactions happen in one ongoing thread rather than across separate conversations, and it pulls in your existing Claude memories, skills, and connected apps, so there's minimal setup.

The proactive part: briefs and background checks

The daily rhythm Anthropic describes:

A short morning brief covering important items across your connected apps — emails, calendar, messages — before your day starts.
Checks every few hours across those same apps, with an alert when something needs attention.
Background work in the cloud — Duet keeps working even when your devices are off, using its own virtual machine and browser rather than needing your computer.
Updates when tasks finish or when something needs your decision.

This is the same "always-on" pitch OpenAI and Meta are making — the race is no longer about the smartest chatbot, but about who gets to be the agent that works while you sleep.

What's under the hood

Built on Claude Opus 5.5 — Anthropic's current flagship-tier model, also listed at $4/$20 per million input/output tokens on its API pricing.
6,000+ connectors through the Claude Marketplace, including Gmail, Slack, QuickBooks, Stripe, Gusto, LinkedIn, DocuSign, HubSpot, and Shopify, per press reporting.
Its own virtual machine and browser to complete tasks independently, without local access to your computer.
Purchases via Stripe Link and Shopify Shop Pay — Duet can make purchases through these digital wallets (with your approval).
Memory, skills, and connected apps carry over from your Claude account, and it can search past conversations.

Safety and approvals

Anthropic's line here is clear and worth quoting: Duet gets user approval before it takes actions like sending a message, making a payment, signing a document, or deleting a file. Additional guardrails per the launch reporting:

Duet does not see passwords or logins when given access to accounts.
A dedicated classifier and safety system prompt checks each message and guards against sensitive information leaks.
Prompts follow structured, validated sequences during sensitive operations rather than free-form model output.

Anthropic was also honest about the weak spots: Duet can misjudge urgency and remains less effective at personal tasks such as booking tickets. That's a refreshing admission — and a reminder that the research-preview label is doing real work here.

Pricing and access

Three things to know about getting in:

Research preview via waitlist. Duet is not generally available — access is granted through a waitlist, and it's initially limited to US subscribers on Pro and Max plans.
Included in your plan. There's no separate Duet price: it uses your existing Claude plan limits. Pro is $20/month ($17/month billed annually); Max is $100/month (5x tier) or $200/month (20x tier), verified against third-party pricing roundups in October 2026. Background checks count toward your usage limits, and complex assignments consume more — so a busy Duet could eat a meaningful share of your allowance.
Web + mobile. Available via standalone iOS and Android apps or on claude.ai — no extra download cost.

Pricing was verified October 2026 via independent reporting; Anthropic changes plans often, so check the official pricing page before you subscribe.

How it stacks up

The proactive-agent race now has three named entries:

OpenAI's Dots — always-on agents that pursue user goals across apps like Slack and Microsoft Teams, launched the month before Duet (September 2026, per Reuters).
Meta's Muse agent — per Reuters, it drew millions of downloads and fueled investor interest in autonomous AI services.
Anthropic's Duet — the newest entrant, leaning on Anthropic's reputation for safety: explicit approval gates, a dedicated safety classifier, and memory/skills portability from the Claude ecosystem.

Compared with Google's agent offerings and Meta's Muse, Duet's distinctive pitch is the approval-first design: it asks before sending, paying, signing, or deleting. In a field where agents act autonomously, "asks first" is itself the differentiator.

What it means for creators

For a one-person creator business, the pitch lands directly: an agent that watches your inbox for brand-deal emails, flags calendar conflicts, drafts replies, and pings you when something needs a decision — while you're filming. The 6,000+ connector library (Shopify, Stripe, HubSpot) reads like it was built with small businesses in mind.

You should care if you: run a business on top of Gmail, Slack, Shopify, or QuickBooks and already pay for Claude Pro or Max — Duet is included in the plan you're already on.

You can safely ignore it if you: are outside the US, on the free tier (no access), or wary of granting an agent ongoing access to your accounts — a research preview with known urgency-misjudgment issues is not the place to start handing over your business's keys.

Verdict

Claude Duet is the most consequential agent launch of October 2026 — not because the concept is new (OpenAI and Meta got there first), but because Anthropic is making the safety-first version: approval gates on every sensitive action, a dedicated safety classifier, and an honest admission that it misjudges urgency. The waitlisted, US-only research preview keeps expectations honest. If you've been curious about always-on agents but nervous about letting one act autonomously, Duet is the one to watch — join the waitlist and let the early users surface the real failure modes first.

FAQ

What is Claude Duet?

A proactive AI agent app Anthropic unveiled on October 21, 2026. It connects to your apps, sends a daily morning brief, checks apps every few hours, and completes tasks in the cloud — reaching out before you ask, rather than waiting for prompts.

How much does Claude Duet cost?

There's no separate price: Duet is included with Claude Pro ($20/month) and Max ($100–$200/month) plans in the US, sharing your existing usage limits. Background checks count toward those limits. Access is via a waitlist for a research preview.

Is Claude Duet available outside the US?

Not at launch. It's initially limited to US subscribers on Pro and Max plans, in a research preview accessed through a waitlist.

What can Claude Duet do without my permission?

It requires your approval for sensitive actions: sending messages, making payments, signing documents, or deleting files. Anthropic says it does not see passwords or logins, and a dedicated safety classifier checks each message.

How is Claude Duet different from OpenAI Dots and Meta Muse?

All three are proactive agents that work across your apps. Duet is built on Claude Opus 5.5, integrates 6,000+ connectors from the Claude Marketplace, and differentiates on approval-first safety design — asking before sensitive actions rather than acting autonomously.

What model powers Claude Duet?

Claude Opus 5.5, Anthropic's current flagship-tier model.