Anthropic launches ‘Cowork’ mode to turn Claude AI into office assistant

Anthropic has rolled out Cowork, a new desktop-focused feature that enables its Claude AI to actively manage files and tasks, marking a shift from chat-based assistance to hands-on workflow automation for everyday office work.

Anthropic has announced the launch of Cowork, a new capability designed to expand the role of its Claude AI assistant beyond conversations and into practical desktop execution. The feature is being offered as a research preview to Claude Max subscribers on macOS, signalling the company’s ambition to make AI a more integrated part of daily work routines.

Unlike traditional chatbot interactions, Cowork allows users to grant Claude controlled access to specific folders on their computers. Once permission is given, the AI can read, edit, organise and create files within those directories. Anthropic says this enables Claude to take on tasks that typically require manual effort, such as sorting downloads, renaming files, compiling spreadsheets from screenshots, or drafting reports from fragmented notes.

The company positions Cowork as a bridge between conversational AI and autonomous task execution. Instead of responding to isolated prompts, Claude can plan a sequence of actions and work through them step by step, while keeping users informed of progress along the way.

From conversations to coordinated workflows

Anthropic says Cowork is built on the same technical foundation as Claude Code, its AI agent aimed at developers. While Claude Code focuses on programming and technical workflows, Cowork adapts those underlying capabilities for non-technical users handling everyday administrative and creative tasks.

Once assigned a job, Claude operates with greater independence than in a standard chat interface. Users can queue multiple tasks, provide feedback mid-process, and allow the AI to complete work in parallel. According to Anthropic, the experience is meant to resemble delegating work to a human colleague rather than managing a rigid back-and-forth conversation.

Safeguards and user control

Despite its increased autonomy, Anthropic emphasises that Cowork is designed with user oversight in mind. Claude is expected to request confirmation before taking significant actions, allowing users to review, redirect or refine tasks as needed. This approach aims to balance efficiency with transparency and trust.

Anthropic says Cowork is intended to reduce the friction involved in preparing context, formatting outputs, and manually transferring AI-generated content into usable files. By embedding the assistant directly into desktop workflows, the company believes Cowork can help users save time while maintaining control over their work.

The research preview phase suggests Anthropic will continue refining Cowork based on user feedback, as it explores how AI agents can evolve from helpful assistants into reliable collaborators across everyday professional tasks.