Anthropic unveils Claude Duet, a proactive AI agent
publishers 145articles 265first reported 24 Sep, 16:43 UTCdeveloping since 24 Sep · 36 editions
Anthropic has announced Claude Duet, a personal AI agent that can monitor connected apps, prepare daily briefings and carry out tasks in the background. The tool is launching as a research preview, with a waitlist initially limited to Claude subscribers in the United States.
Duet can handle administrative work such as scheduling meetings as well as more complex assignments, including creating forecasts, dashboards, prototypes and web apps, Anthropic said. It runs on the company’s Opus 5.5 model in its own virtual machine, allowing it to continue working without an active Claude session. Standalone Android and iOS apps are also planned.
Anthropic said Duet will seek confirmation before acting on a user’s behalf and will send messages only when attention is needed. Users will be able to control how frequently it contacts them and which topics trigger notifications. The agent will check connected accounts every few hours rather than continuously to conserve plan usage.
The company said Duet includes safeguards intended to prevent disclosure of sensitive data. It will not directly access passwords or sign-in tokens, while health or financial information can be shared only with explicit user authorization. Anthropic said further details about its safety protections would follow.
When Duet becomes generally available, Anthropic plans to offer it to subscribers on its $20-per-month Pro plan.
HOW THIS STORY WAS MADE
Written from 8 articles, 3 with the publisher's own text; 8 independent newsrooms once syndicated copies count as one; this version written 558 h after the record first saw the story; the editor revised it.
- Publishers
- 145
- Source articles
- 265
- Given to the writer
- 8, 3 with the article's own text
- Independent newsrooms
- 8
- This version written
- 558 h after the record first saw the story
- Second model (editor)
- revised
- Publication gate
- passed
- Human review
- none
WHAT THE EDITOR DID
The second model checked the writer's draft against the same sources and changed it before publication. Struck words were in the draft; underlined words are its corrections.
WHAT IT FLAGGED editor revise: Delete the unsupported statement about the absence of a general release date. - "The company has not announced a general release date.": The source does not report that no general release date has been announced.
THE DRAFT AND THE CHANGES
Anthropic unveils Claude Duet, a proactive AI agent
Anthropic has announced Claude Duet, a personal AI agent that can monitor connected apps, prepare daily briefings and carry out tasks in the background. The tool is launching as a research preview, with a waitlist initially limited to Claude subscribers in the United States.
Duet can handle administrative work such as scheduling meetings as well as more complex assignments, including creating forecasts, dashboards, prototypes and web apps, Anthropic said. It runs on the company’s Opus 5.5 model in its own virtual machine, allowing it to continue working without an active Claude session. Standalone Android and iOS apps are also planned.
Anthropic said Duet will seek confirmation before acting on a user’s behalf and will send messages only when attention is needed. Users will be able to control how frequently it contacts them and which topics trigger notifications. The agent will check connected accounts every few hours rather than continuously to conserve plan usage.
The company said Duet includes safeguards intended to prevent disclosure of sensitive data. It will not directly access passwords or sign-in tokens, while health or financial information can be shared only with explicit user authorization. Anthropic said further details about its safety protections would follow.
When Duet becomes generally available, Anthropic plans to offer it to subscribers on its $20-per-month Pro plan. The company has not announced a general release date.
Written by a language model from the sources above, then checked by a second model that may only cut, attribute or correct. How it works →