What happened
Anthropic launched Claude Duet, a mobile and desktop designed to proactively monitor connected applications, surface priorities, and execute background tasks. The product, built on the Opus 5.5 model, is currently in a research preview for U.S. subscribers on Pro and Max plans via a waitlist. It operates through a dedicated app and claude.ai, using a remote computing environment to continue working after devices are closed. Key features include daily briefings, invoice tracking, and draft generation, with explicit approval required for consequential actions like payments or deletions.
Anthropic introduced Claude Duet on Wednesday, an AI assistant that monitors connected applications and carries out assignments in the background. The product is built around the Opus 5.5 model and is available as a research preview to U.S. subscribers on Pro and Max plans through a waitlist. It operates via a separate mobile app for iOS and Android and through claude.ai, utilizing a remote computing environment that allows tasks to continue after a user closes their laptop or turns off their phone.
The agent’s core function is proactive monitoring. It generates daily briefings from connected services and checks accounts periodically for developments warranting notification. Users can adjust which subjects are flagged and the frequency of updates, though Anthropic notes that more frequent monitoring consumes usage allowances faster. Duet runs in a virtual machine with a browser, determining the necessary applications and steps for requested assignments and notifying users upon completion or when a decision is needed.
Anthropic emphasizes a boundary between preparing work and taking consequential action. Duet can inspect information and assemble drafts independently, but it requires explicit user approval before transmitting messages, making payments, signing agreements, or deleting material. Certain activities, such as securities transactions and transfers between financial accounts, remain prohibited even with authorization. The system includes a sensitive-information classifier to block the sharing of restricted data like payment-card numbers and government-issued identifiers.
The launch includes integrations with over 6,000 connectors and plugins from the Claude Marketplace. Announced workflows cover business functions such as examining QuickBooks or Stripe records for unpaid invoices, drafting customer outreach using HubSpot data, and managing Gusto payroll. Duet also supports purchases through Stripe’s Link wallet and Shopify’s Shop Pay, using virtual cards to protect underlying card numbers. A forthcoming feature will allow Duet to use credentials stored in 1Password with user permission.
Anthropic states that internal security teams and an outside firm tested Duet’s infrastructure, and the product is covered by its public bug bounty. The company acknowledges that the preview can misunderstand conversations and assign incorrect urgency to issues. It also notes that Duet cannot yet display its browser to the user, limiting visibility into its work. The product is included in existing Pro and Max subscriptions, with no separate charge announced, but automated checks consume part of the user’s existing usage allowance.
Source details: venturebeat.com ↗
Why it matters
Claude Duet represents a significant shift in design from reactive chatbots to proactive operational assistants that initiate work without user prompts. By integrating with over 6,000 connectors and handling multi-step business workflows, it targets a gap in current AI offerings where users must manually initiate every interaction. This launch intensifies competition with Meta’s Muse and OpenAI’s Dots, forcing a reevaluation of how AI agents manage permissions, privacy, and autonomous decision-making in professional settings. The product’s reliance on existing subscription allowances rather than separate pricing also changes the cost calculus for businesses adopting agentic AI.
Claude Duet marks a strategic pivot for Anthropic toward proactive, autonomous agents that initiate work rather than waiting for user prompts. This addresses a key limitation in current AI assistants, where users must manually trigger every interaction. By monitoring connected accounts and surfacing priorities, Duet aims to reduce the cognitive load on professionals and small business owners who manage multiple systems.
The product’s architecture, which uses a remote computing environment, allows for continuous background processing. This is a significant technical distinction from earlier features like Claude Code Remote Control or Dispatch, which rely on local devices remaining active. This shift enables more reliable execution of time-sensitive or long-running tasks, such as financial forecasting or customer follow-ups, without requiring constant user supervision.
The launch intensifies competition in the emerging market for persistent personal and business agents. Meta’s Muse and OpenAI’s Dots offer similar proactive capabilities, but Duet’s integration with a broad ecosystem of business tools and its explicit approval workflows for consequential actions position it as a serious contender for enterprise and professional use. The lack of a separate fee for Duet, bundled within existing subscriptions, lowers the barrier to entry for users already paying for Claude Pro or Max.
However, the product’s reliance on shared usage allowances introduces a new cost consideration. Automated monitoring consumes part of the user’s existing quota, meaning that heavy use of Duet’s background checks could reduce the capacity for other AI tasks. This creates a budgeting challenge for businesses that need to balance proactive monitoring with other AI-driven workflows, and it highlights the need for transparent cost management in agentic AI systems.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
crm_get_transaction(id='4092').An agent must create a draft calendar event for Tuesday at 2 p.m. Which evidence would establish the requested result?
What to watch next
Monitor the expansion of Claude Duet to Team and Enterprise subscriptions, which are currently excluded from the preview. Watch for the release of the Duet API and the implementation of 1Password credential integration. Additionally, track independent evaluations of the agent’s accuracy in multi-system workflows and its handling of risks, as Anthropic’s internal testing claims have not been externally verified.
The current preview is limited to U.S. subscribers on Pro and Max plans, with no access for Team or Enterprise subscriptions. Watch for Anthropic’s announcement of an enterprise rollout schedule and the availability of organization-level features such as audit logs and custom retention controls, which are currently listed under Enterprise but not confirmed for Duet.
Anthropic has not announced a Duet API, which limits third-party integration and development. The release of an API would be a significant step toward making Duet a platform for building custom agentic workflows, and its timing will be a key indicator of Anthropic’s long-term strategy for the product.
The forthcoming integration with 1Password for credential management is a critical security feature that will determine how safely Duet can handle sensitive authentication. The implementation details and security guarantees of this feature will be closely scrutinized by enterprise buyers and security experts.
Independent evaluations of Duet’s accuracy and reliability in multi-system workflows are currently lacking. Anthropic’s internal testing claims have not been externally verified, and the company acknowledges that the preview can misunderstand conversations. Watch for third-party benchmarks and user reports that assess the agent’s performance in real-world business scenarios.