Anthropic launches Claude Duet, a proactive AI agent for $20/month

Anthropic has officially launched Claude Duet, a new personal AI agent designed to proactively manage daily tasks and communicate only when necessary. The tool marks the company’s entry into the crowded market of autonomous digital assistants, directly competing with recent releases from rival tech giants. Unlike previous iterations of AI chatbots that wait for user input, Duet is built to operate independently, handling administrative duties and complex projects in the background. The launch represents a significant strategic shift for Anthropic, moving beyond passive conversation tools to active, self-directed agents that can execute multi-step workflows on behalf of the user.

The move positions Anthropic as a late but aggressive competitor to OpenAI’s Dots and Meta’s Muse. While those competitors have already established a foothold in the personal agent space, Anthropic argues that Duet offers a more balanced approach to autonomy. The system is powered by the company’s Opus 5.5 model and runs on a dedicated virtual machine, meaning users do not need to keep a Claude session open in their browser. This architecture allows the agent to work continuously, generating daily briefings and executing tasks ranging from scheduling meetings to building web applications. By separating the agent’s execution environment from the user’s active interface, Anthropic aims to reduce friction and allow the AI to handle work while the user focuses on other activities.

A key differentiator for Duet is its pricing and accessibility model. While OpenAI currently restricts its Dots agent to its high-tier Pro subscribers, which cost at least $100 per month, Anthropic plans to include Duet in its standard $20 per month Pro plan. This makes the technology accessible to a much broader audience of individual professionals and hobbyists. However, Anthropic acknowledges potential challenges with resource usage. Because Duet shares usage limits with other Claude tools like Code, there is a risk that the agent’s background checks could deplete a user’s monthly allowance quickly. To mitigate this, the company has implemented a throttling mechanism where Duet checks connected accounts only every few hours rather than continuously. Users will also have control over notification frequency and topics, allowing them to tailor the agent’s behavior to their specific needs.

Security and privacy remain central concerns as users are invited to connect sensitive personal and professional data to an autonomous agent. Anthropic has introduced a new classifier and specific system prompts designed to prevent the leakage of sensitive information. The agent is engineered to never directly access passwords or sign-in tokens. Furthermore, data related to health or personal finances will only be processed if the user explicitly authorizes the sharing. Despite these safeguards, the ultimate success of Duet will depend on user trust. For many, granting an AI agent such deep access to their digital life requires a high level of confidence in the provider’s security infrastructure. As the waitlist opens for US-based subscribers, the coming weeks will provide early insights into whether Duet can balance autonomy with safety effectively.