AI · 2026-10-22
Anthropic Unveils “Claude Duet” — an AI Agent That Looks After Work First
Anthropic has introduced “Claude Duet,” an AI agent that checks schedules, messages, and connected work tools to find what needs to be done and carry it out. Major actions are designed to go through user approval, and the service is initially available as a research preview to Claude Pro and Max subscribers in the United States.

Artificial intelligence is moving beyond simply answering questions or drafting documents and is evolving toward a model that identifies tasks on its own and gets to work first. Anthropic’s standalone AI agent application, “Claude Duet,” is designed to review connected calendars, messages, and work information so it can organize what needs attention and carry out tasks without requiring the user to enter instructions every time.
According to a report by Jemin Ilbo, Anthropic unveiled Claude Duet on October 21, 2026. The service can summarize key appointments and work status each morning, then check connected apps throughout the day and send a notification when something requires the user’s judgment. Unlike a conventional chatbot that waits for questions in a chat window, it continues working in the background and notifies the user when results are ready or a decision is needed.
### From schedule management to business analysis
The tasks Claude Duet is designed to support go beyond everyday work such as coordinating schedules. They also include building dashboards with multiple tools, analyzing financial and sales forecasts, developing prototypes, and creating web applications. Anthropic also highlighted use cases such as checking email or work messengers to draft replies, and reviewing overdue payments to prepare payment notices—work that connects several different business tools.
Anthropic said the service is powered by its own Claude Opus 5.5 model and can use more than 6,000 integrations and plugins available through the Claude Marketplace. This setup can be seen as an effort to evolve AI from a helper for individual tasks into a collaborator that connects multiple apps and workflows.
### Between automation and user control
When it comes to AI that carries out work on behalf of users, convenience is not the only concern; maintaining user control is just as important. Claude Duet is designed to show users the details and obtain approval before taking high-impact actions such as sending messages, making payments, signing documents, or deleting materials. The company also says it has safeguards to prevent the service from directly checking passwords or login credentials, and to reduce the external sharing of sensitive personal information.
The service has begun rolling out as a research preview through a waitlist for Claude Pro and Max subscribers in the United States. Initially, access will be granted gradually to selected users, and the amount of background work used will count toward the existing usage limits of the standard Claude service. In other words, if you delegate too many tasks, the limit for other features may be reached more quickly. Users can also adjust the topics and frequency of notifications.
### The next benchmark in the AI agent race
Major companies such as OpenAI and Meta are also investing heavily in AI agents that move across multiple applications to help users achieve their goals. In this competition, product success may no longer be decided solely by how well a model generates answers. Equally important are whether it understands work context, whether it can reliably use connected tools, and whether there are safeguards to prevent mistaken actions.
How much time Claude Duet actually saves in real-world work, and how much it reduces errors, will depend on user feedback and future feature improvements. In particular, adoption will hinge on whether people are willing to connect sensitive data such as email, calendars, and business information to AI, and whether the usage limits and approval process feel sufficient. In an environment where AI acts before being asked, trust and control may matter just as much as convenience.
According to a report by Jemin Ilbo, Anthropic unveiled Claude Duet on October 21, 2026. The service can summarize key appointments and work status each morning, then check connected apps throughout the day and send a notification when something requires the user’s judgment. Unlike a conventional chatbot that waits for questions in a chat window, it continues working in the background and notifies the user when results are ready or a decision is needed.
### From schedule management to business analysis
The tasks Claude Duet is designed to support go beyond everyday work such as coordinating schedules. They also include building dashboards with multiple tools, analyzing financial and sales forecasts, developing prototypes, and creating web applications. Anthropic also highlighted use cases such as checking email or work messengers to draft replies, and reviewing overdue payments to prepare payment notices—work that connects several different business tools.
Anthropic said the service is powered by its own Claude Opus 5.5 model and can use more than 6,000 integrations and plugins available through the Claude Marketplace. This setup can be seen as an effort to evolve AI from a helper for individual tasks into a collaborator that connects multiple apps and workflows.
### Between automation and user control
When it comes to AI that carries out work on behalf of users, convenience is not the only concern; maintaining user control is just as important. Claude Duet is designed to show users the details and obtain approval before taking high-impact actions such as sending messages, making payments, signing documents, or deleting materials. The company also says it has safeguards to prevent the service from directly checking passwords or login credentials, and to reduce the external sharing of sensitive personal information.
The service has begun rolling out as a research preview through a waitlist for Claude Pro and Max subscribers in the United States. Initially, access will be granted gradually to selected users, and the amount of background work used will count toward the existing usage limits of the standard Claude service. In other words, if you delegate too many tasks, the limit for other features may be reached more quickly. Users can also adjust the topics and frequency of notifications.
### The next benchmark in the AI agent race
Major companies such as OpenAI and Meta are also investing heavily in AI agents that move across multiple applications to help users achieve their goals. In this competition, product success may no longer be decided solely by how well a model generates answers. Equally important are whether it understands work context, whether it can reliably use connected tools, and whether there are safeguards to prevent mistaken actions.
How much time Claude Duet actually saves in real-world work, and how much it reduces errors, will depend on user feedback and future feature improvements. In particular, adoption will hinge on whether people are willing to connect sensitive data such as email, calendars, and business information to AI, and whether the usage limits and approval process feel sufficient. In an environment where AI acts before being asked, trust and control may matter just as much as convenience.
Related search terms
#Anthropic Claude Duet#Claude Duet AI agent#Anthropic AI agent#Claude Pro Max research preview#AI workflow automation#schedule and message management AI#Claude Marketplace integrations