Claude Duet arrives October 21: Anthropic's $20 answer to OpenAI's $100 agents

Anthropic released Claude Duet on October 21, becoming the latest AI lab — after SpaceXSI, Meta and OpenAI — to ship a personal agent, and it did so at a price point that undercuts its closest rival by eighty percent: $20 per month. The agent is available in preview to Claude Pro and Max subscribers in the US, and only after they join a waitlist on Anthropic's website. The launch matters because it marks the moment the major AI labs are converging on the same product thesis — always-on software that acts on your behalf rather than responding when prompted — and Anthropic is the first to bundle it into a mainstream consumer tier instead of a premium one.

What is Claude Duet?

Claude Duet is an always-on AI agent that Anthropic bills as capable of proactively completing tasks for you even when you are away from your phone or computer. That proactivity is the defining difference from the standard Claude experience, and Anthropic articulated it in unusually plain terms.

"The Claude app waits for you to open a chat, and you work on something alongside Claude. Claude Duet comes to you," Anthropic says. "It tells you what needs your attention each morning, takes tasks you hand it by message, and reaches out when one is done or needs your approval."

The inversion is structural, not cosmetic. Duet runs on its own virtual machine (a software-simulated computer operating independently of your device), with access to a dedicated browser it can use to surf the web and complete jobs. It can also use any apps you have connected to your Claude account.

Each day, the agent generates a short briefing with a list of action items based on those connected apps. Users can delegate tasks to it directly. Anthropic's examples include:

• Scheduling meetings
• Creating forecasts
• Building web apps
• Generating a weekly sales and cashflow dashboard, in the case of a business owner
• Flagging newly published research papers relevant to a scientist's field of study

The breadth of those examples signals Anthropic's ambition: Duet is aimed simultaneously at knowledge workers, operators of small businesses and researchers, not at a single vertical.

How do you get access?

Availability is narrow at launch. Duet is in preview, limited to Claude Pro and Max subscribers in the United States, and gated behind a waitlist hosted on Anthropic's website. Joining it requires more than an email address. Anthropic asks for:

• Your name
• Your company
• Your email address
• The type of phone you use — Android or iPhone
• A short description of the work you intend to hand off to Claude

The company asks about your device because it offers a standalone Claude Duet app on both Android and iOS, which means the agent is designed to live on your phone rather than only inside a browser tab.

What does it cost compared with the competition?

Anthropic says it will include Duet as a perk of its $20 per month Pro plan and up. The company notes this could change, but the positioning is deliberate. OpenAI's comparable Dot agents are currently available only through OpenAI's own Pro plans, which start at $100 per month — and OpenAI recently added a $500 per month option on top of that.

Five times the entry price. Twenty-five times at the top tier. That gap frames the competitive stakes of this launch: Anthropic is treating the agent not as a luxury feature but as part of the standard subscription, at least for individual users. There is one significant carve-out — Anthropic is not making Claude Duet available to Team and Enterprise customers just yet, which leaves business deployments, arguably the most natural market for delegated dashboards and forecasting, waiting.

Why does Duet only check your accounts every few hours?

Usage economics. During the preview, Anthropic says Duet will share usage with all your other Claude apps, including Code and Cowork. That pooling means the agent draws from the same allowance that subscribers already spend on chat and coding tools.

Anthropic has said it designed the agent to check your connected accounts every few hours, "because checking nonstop would use up your plan's limits much faster." The cadence is a direct concession to the cost of running an always-on agent against metered plan limits.

Part of the reason the company is previewing Duet before a wider rollout is to collect feedback on how often it messages users. Frequency of interruption is evidently an open design question, not a settled one. Reviewers have not yet established whether the agent is genuinely usable within a Pro plan's limits, and that question will likely determine whether the $20 positioning holds up in practice.

What is the privacy tradeoff?

Duet's proactivity is built on the data in your Claude account. That includes information the chatbot has written to its memory and any apps connected to your Anthropic account. The agent's morning briefings, its awareness of what "needs your attention" and its ability to act all depend on that data foundation.

The consequence is straightforward: how useful you find Duet will depend heavily on how much of your digital life you are willing to share with Anthropic. Connecting calendars, documents and work apps gives the agent raw material; withholding them hollows it out. For some users, that privacy tradeoff will not be worth it, a calculation that predates Duet but that always-on agents sharpen considerably, because the data flows continuously rather than session by session.

What happens next?

The preview will tell. Anthropic is using this phase to gauge messaging frequency, observe real-world task delegation and decide when to widen availability beyond US Pro and Max subscribers — and eventually, presumably, to the Team and Enterprise customers currently excluded. The open questions are whether an every-few-hours agent on a shared $20 plan can deliver enough completed tasks to justify the data it consumes, and whether OpenAI responds by unbundling Dot from its $100 floor. The answer will shape how quickly always-on agents move from premium experiment to default feature of every AI subscription.