Anthropic, a leading AI technology company, has announced new capabilities for its Claude AI tools, Code and Cowork, allowing users to control their computers autonomously. This update enables tasks like opening files, using web browsers and apps, and running dev tools without any setup, even when the user is away from their computer.
The enhanced Claude AI functionality is currently available as a research preview for Claude Pro and Max subscribers, exclusively on macOS devices. By leveraging autonomous capabilities introduced in Claude’s 3.5 Sonnet AI model in 2024, the update now extends these capabilities to the Code and Cowork AI agents tailored for programmers.
To utilize this feature, users need to have the Claude desktop app running on a compatible macOS device and paired with the chatbot’s mobile app. The system prioritizes connectors to supported services such as Slack and Google Workspace apps, resorting to direct control of the browser, mouse, keyboard, and display if a connector is unavailable. Anthropic emphasizes that Claude always seeks explicit permission before executing actions like exploring, scrolling, and clicking to complete tasks.
According to Anthropic, this feature synergizes particularly well with Dispatch, a recent cross-device functionality enabling task assignment to the Claude desktop app from a mobile device. While the company acknowledges that complex tasks may require a second attempt and that screen-based interactions are slower than direct integrations, they are releasing this feature early to gather feedback on its performance and limitations.
Source: The Verge