Claude Duet pricing, plans and limits

Best for founders, small business owners and busy professionals who already pay for Claude and want a chief of staff that chases loose ends for them. It replaces the daily sweep through inbox, calendar and billing tools, not your project software. One early tester, a kebab restaurant owner with about 110 employees, used it for accounting, support and marketing.

Claude Duet is a proactive AI agent that Anthropic released as a research preview on October 21, 2026. It runs in its own app on a cloud virtual machine, watches the apps you connect, and completes delegated work, but it stops for your approval before sending messages, making payments, signing documents or deleting anything.

Maker: Anthropic · Autonomy: semi autonomous · Maturity: ALPHA

Underlying models: Claude Opus 5.5

About Claude Duet

Claude Duet is a separate app from Anthropic in which Claude works as a proactive agent: instead of waiting for a prompt, it comes to you with what needs attention and keeps working when you step away. Anthropic opened it as a research preview in October 2026, starting with a waitlist for US subscribers, and says it built Duet for professionals and small business owners who have more on their plates than they can track. The difference from the regular Claude app is the direction of travel. Claude waits for you to open a chat; Duet reaches out to you.

Anthropic designed Duet around Claude Opus 5.5 and tuned it to decide what needs you, what can wait, and when to check with you first. Everything happens in one running thread that you type or speak into, whether the request takes a minute or a week. Describe a job in a short message and it works out which connected apps and which steps it needs, then runs them in the background. The agent tracks your accounts on a schedule rather than constantly, which keeps its own consumption of your plan in check.

The launch material leans toward running a business rather than personal errands. Anthropic's examples include chasing unpaid invoices, preparing payroll for review, getting contracts ready to sign and turning a product launch into a social campaign. Inside Anthropic, the lead of the Claude Code team talks through a mental to-do list with it on walks, and the lead of the Duet team uses it to schedule user interviews, work on prototypes and ship code.

Duet lands in a category that filled up fast in 2026. OpenAI dots live inside ChatGPT, Muse from Meta targets shops that sell through Instagram and Facebook, xAI offers Grok Bot, and Poke works from your text messages. Our comparison of Grok Bot, Muse and dots tests those first three against each other. Self-hosters can run the open-source OpenClaw or Hermes Agent on their own hardware, and developers who mainly want code written are better served by Claude Code. More options sit in the general-purpose agents section.

Anthropic says what it learns from early users will decide how Duet and the main Claude app eventually work together. It has named two additions as coming soon: signing in to websites with logins you have saved in 1Password, with your permission, and a richer visual editor for designing a Shopify storefront and previewing it before anything goes live. It has also promised to publish more detail on Duet's safety design in the days after launch.

Pricing

Claude Duet has no price of its own. During the research preview it comes with a paid [Claude](/hub/tools/claude) subscription for US users who clear the waitlist. On Anthropic's pricing page, Pro costs $20 a month, or $17 a month on annual billing with $200 charged up front, and Max from $100 a month with a choice of 5x or 20x the Pro allowance, billed monthly only. The page does not print a separate figure for the 20x tier. Duet shares one usage pool with your chats and Claude Code, and its scheduled checks of your accounts draw from that pool.

Key Features

• Morning brief: Each morning it sends a short brief on the most important items across your connected apps, then messages you during the day only when something needs a decision.
• Runs on its own cloud machine: Each job runs on a virtual machine with its own browser, so work continues with your laptop and phone switched off, and a notification arrives when a task finishes or needs review.
• Approval gate with hard limits: Before it sends a message, makes a payment, signs a document or deletes anything, it shows the exact action and waits; trading securities and moving money between accounts stay blocked even if you approve.
• Over 6,000 connectors and plugins: It draws on Anthropic's Claude Marketplace, with named examples including Gmail, Slack, Intuit QuickBooks, Stripe, Gusto, HubSpot, Docusign, Canva and Shopify.
• Purchases without sharing your card: Purchases go through Link, Stripe's digital wallet, which asks you to approve the spend and pays with a virtual card, or through Shop Pay at online stores.
• Sensitive-information classifier: A dedicated classifier and safety system prompt check each message, are designed to refuse sharing card numbers or government ID, and pass health or personal-finance details only on your explicit instruction.
• Builds finished work products: Anthropic lists dashboards built from dozens of tools, financial and sales forecasts, decks and web app prototypes among the larger tasks it can take on.

Strengths

• Every's Dan Shipper, after two weeks of early access, called it the first personal agent worth describing as a professional tool (Anthropic gave Every access but had no input on the review).
• No extra subscription: Pro and Max subscribers get it at no added charge during the preview.
• Its credential handling follows Anthropic's Managed Agents design, so it never directly sees your passwords or sign-in tokens, and red teams plus an outside firm have penetration tested it.
• Per Every's review, it splits recurring topics into their own channels and keeps a running list of what it is doing and what it needs from you.

Weaknesses

• Anthropic warns it can misread a thread, misjudge what is urgent, and miss something that arrives between its check-ins, which happen every few hours.
• Personal errands such as booking tickets are weaker than work tasks, and you cannot yet watch its browser while it works.
• In Every's test, scheduled runs could not reach custom connectors, it could see but not direct Claude sessions on the reviewer's computers, and voice input was transcription rather than a full voice mode.
• Background checks and large jobs spend the same allowance as your chats and coding, so heavy use can push a Pro subscriber toward Max.

Visit Claude Duet Official Site