Claude Cowork: Your AI-Powered Knowledge Work Partner

Claude Cowork from Anthropic transforms how you handle everyday knowledge work tasks. This autonomous AI agent reads, edits, and creates files in your workspace, turning hours of manual work into minutes of intelligent automation.

Download Claude Desktop for macOS

Available for Claude Max subscribers • macOS only

New: Launched January 12, 2026 by Anthropic as a research preview

What is Claude Cowork?

Claude Cowork is Anthropic's breakthrough extension of Claude Code, specifically designed for knowledge workers who need autonomous file management and content creation capabilities without writing a single line of code. Built on the same Claude Agent SDK that powers Claude Code, Claude Cowork represents Anthropic's commitment to making advanced AI accessible for everyday productivity tasks.

Following the success of Claude Code—Anthropic's AI coding assistant—the team at Anthropic observed users creatively adapting the tool for non-technical knowledge work. This insight led to the rapid development of Claude Cowork, which was built in approximately 10 days largely using Claude Code itself, demonstrating the power of AI-assisted development.

As a research preview launched January 12, 2026, Claude Cowork is available exclusively to Claude Max subscribers on macOS Desktop, with Windows support planned for future releases. The tool operates within a sandboxed virtual machine environment, ensuring security while providing powerful autonomous capabilities for file manipulation and content generation.

Developed by Anthropic, the creators of Claude—one of the world's most capable AI assistants trusted by enterprises and individuals worldwide.

Powerful Capabilities for Knowledge Work

📊

Autonomous File Operations

Claude Cowork reads, edits, and creates files independently within your designated workspace folder. Grant folder-level permissions once, and the agent handles all file operations autonomously while keeping you informed of progress.

Parallel Task Execution

Using sub-agent architecture, Claude Cowork breaks complex tasks into smaller components and executes them in parallel, dramatically reducing completion time for multi-file operations and research-intensive projects.

🔒

Sandboxed Security

All operations run in a containerized virtual machine environment, isolating Claude Cowork from your broader system. Access is limited strictly to folders you explicitly grant, ensuring data security and system integrity.

📝

Intelligent Content Creation

From expense reports to presentation decks, Claude Cowork analyzes source materials and generates professional-quality documents, maintaining consistent formatting and extracting relevant information accurately.

🗂️

Smart File Organization

Transform chaotic folders into organized systems. Claude Cowork analyzes file contents, renames files meaningfully, creates logical folder structures, and maintains organization schemes that make sense for your workflow.

🔍

Research Synthesis

Aggregate scattered notes, compile research findings, and synthesize information from multiple sources into coherent reports. Claude Cowork extracts key insights and presents them in clear, structured formats.

How Claude Cowork Works

Claude Cowork simplifies knowledge work through a straightforward four-step process that puts AI autonomy to work for you.

01

Grant Folder Access

Open Claude Desktop and navigate to the Cowork interface. Select the specific folder where you want Claude Cowork to operate. The agent can only access files within folders you explicitly grant—ensuring complete control over your data. This folder becomes the workspace where Claude Cowork reads existing files and creates new ones.

02

Describe Your Task

Tell Claude Cowork what you need in natural language. No technical expertise required—simply describe the outcome you want. Examples: "Create an expense report from these receipt screenshots," "Organize my downloads folder by file type," or "Draft a summary report from my meeting notes." Claude Cowork understands context and intent.

03

Watch Autonomous Execution

Claude Cowork analyzes your request, creates an execution plan, and begins working autonomously. The interface displays real-time progress updates as the agent reads files, processes information, and creates outputs. Sub-agents handle parallel tasks when beneficial, optimizing for speed while maintaining quality. You stay informed without needing to micromanage.

04

Review and Refine

When complete, review the results directly in your workspace folder. Claude Cowork creates new files, updates existing ones, or reorganizes structures as requested. Need adjustments? Provide feedback in natural language and Claude Cowork will refine the output. The agent learns from your preferences to improve future task execution.

Real-World Use Cases

Claude Cowork excels at practical knowledge work tasks that typically consume significant time. Here's how professionals are using it.

Expense Report Creation

Finance & Administration

Scenario: You return from a business trip with 15 receipt screenshots in various formats—restaurant bills, taxi receipts, hotel invoices—all needing compilation into a formal expense report.

How Claude Cowork Helps: Point Claude Cowork to your receipts folder and request an expense report. The agent extracts transaction details (date, merchant, amount, category) from each image, creates a formatted spreadsheet with itemized expenses, calculates totals by category, and generates a summary report—all in minutes instead of hours.

Research Synthesis

Knowledge Management

Scenario: You've collected research notes across multiple documents, PDFs, and text files for a market analysis project. Information is scattered and needs consolidation into a coherent report.

How Claude Cowork Helps: Claude Cowork reads all source documents, identifies key themes and insights, extracts relevant data points, and synthesizes findings into a structured report with sections, citations, and executive summary. The agent handles cross-referencing and maintains logical flow throughout the document.

File Organization & Cleanup

Productivity

Scenario: Your Downloads folder contains hundreds of files with generic names making it impossible to find anything quickly.

How Claude Cowork Helps: Claude Cowork analyzes file contents, renames files with descriptive names based on content, creates logical folder structures by type or project, moves files to appropriate locations, and can even suggest deletions for duplicate or outdated files. Your workspace becomes instantly navigable.

Presentation Deck Creation

Business Communication

Scenario: You need to create a presentation deck from meeting notes, existing documents, and data files, but formatting and structuring takes hours away from actual content refinement.

How Claude Cowork Helps: Provide Claude Cowork with your source materials and presentation objectives. The agent creates a structured outline, generates slide content with appropriate headings and bullet points, suggests visual hierarchy, and produces a draft deck ready for your review and customization—dramatically reducing preparation time.

Technical Requirements & Compatibility

System Requirements

  • Operating System: macOS (Windows support planned)
  • Application: Claude Desktop (latest version)
  • Storage: Minimum 2GB free space for VM environment
  • Memory: 8GB RAM recommended for optimal performance
  • Processor: Intel or Apple Silicon Mac

Account & Subscription

  • Subscription Tier: Claude Max required ($100-$200/month)
  • Status: Research preview (early access)
  • Account: Active Anthropic account with Max subscription
  • Availability: Currently macOS Desktop only

Permissions & Access

  • Folder Access: Explicit folder-level permissions required
  • Sandbox: Runs in isolated VM environment
  • File Operations: Read, write, create within granted folders
  • Network: Internet connection required for AI processing

Current Limitations

  • Session Persistence: No session memory between uses (planned)
  • Cross-Device Sync: Not available (single device only)
  • Windows Support: Coming in future release
  • Mobile Access: Desktop only (no mobile version)

Trust, Safety & Privacy

Anthropic built Claude Cowork with security and privacy as foundational principles. Understanding how your data is protected helps you use the tool confidently.

Sandboxed Execution Environment

Claude Cowork operates entirely within a containerized virtual machine, isolating its operations from your broader system. This architecture ensures that even in unlikely error scenarios, the agent cannot access files outside explicitly granted folders or affect system-level configurations.

Explicit Permission Model

You maintain complete control over data access. Claude Cowork can only operate on folders you explicitly grant permission to access. Permissions are folder-specific and can be revoked at any time. The agent never accesses files, folders, or system resources without explicit user authorization.

Transparent Operations

Real-time progress updates show exactly what Claude Cowork is doing at each step—which files it's reading, what changes it's making, and what new content it's creating. This transparency allows you to monitor operations and intervene if needed, maintaining human oversight throughout.

Data Processing Standards

As an Anthropic product, Claude Cowork adheres to the same rigorous data handling standards as all Claude products. File contents processed by the agent are subject to Anthropic's privacy policy and security practices, which prioritize user data protection and confidentiality.

Research Preview Status

Claude Cowork is currently in research preview, meaning Anthropic is actively gathering feedback and refining the system based on real-world usage. This status provides early access to cutting-edge capabilities while allowing Anthropic to ensure the tool meets high standards for production readiness.

Built by Anthropic: Claude Cowork is developed by Anthropic, the AI safety company founded by former OpenAI research executives. Anthropic's mission centers on building reliable, interpretable, and steerable AI systems—principles reflected throughout Claude Cowork's design.

Start Using Claude Cowork Today

Transform your knowledge work with Claude Cowork. Download Claude Desktop for macOS and activate Cowork in your Claude Max subscription to experience autonomous AI assistance for everyday tasks.

Step 1

Subscribe to Claude Max ($100-$200/month)

Step 2

Download Claude Desktop for macOS

Step 3

Access Cowork from the Claude Desktop interface

Step 4

Grant folder permissions and start your first task

Currently available for macOS only. Windows support coming soon. Requires active Claude Max subscription.

Frequently Asked Questions

What is Claude Cowork and how does it differ from Claude Code?

Claude Cowork is Anthropic's autonomous AI agent designed specifically for knowledge work tasks like creating expense reports, organizing files, synthesizing research, and building presentations. While Claude Code focuses on software development and coding tasks, Claude Cowork targets non-technical productivity work. Both are built on the same Claude Agent SDK, but Claude Cowork requires no coding knowledge and operates through natural language instructions for everyday business tasks.

Who can access Claude Cowork?

Claude Cowork is currently available as a research preview exclusively to Claude Max subscribers on macOS Desktop. Claude Max is Anthropic's power-user subscription tier priced between $100-$200 per month. Windows support is planned for future releases, but there is no current timeline for availability on other platforms.

Is my data safe when using Claude Cowork?

Yes. Claude Cowork operates in a sandboxed virtual machine environment, isolating it from your broader system. The agent can only access folders you explicitly grant permission to, and you can revoke access at any time. All file processing follows Anthropic's privacy policy and security standards. Real-time progress updates show exactly what operations the agent is performing, maintaining full transparency throughout task execution.

What types of tasks can Claude Cowork handle?

Claude Cowork excels at knowledge work tasks including: creating expense reports from receipt images, organizing and renaming files in messy folders, synthesizing research notes into coherent reports, generating presentation decks from source materials, extracting and structuring data from documents, and creating summaries from meeting notes. The agent works autonomously on any task involving reading, editing, or creating files within your granted workspace folders.

When was Claude Cowork launched?

Anthropic launched Claude Cowork as a research preview on January 12, 2026. The development team built the entire feature in approximately 10 days, largely using Claude Code itself. This launch followed observations that Claude Code users were creatively adapting the coding tool for non-technical knowledge work tasks, inspiring Anthropic to create a dedicated solution.

Does Claude Cowork require an internet connection?

Yes, Claude Cowork requires an active internet connection to function. While file operations occur locally within the sandboxed environment, the AI processing that analyzes your requests, plans task execution, and generates content happens on Anthropic's servers. This architecture ensures you benefit from Claude's most advanced capabilities while maintaining the security of sandboxed local file access.

Can I use Claude Cowork across multiple devices?

Currently, Claude Cowork does not support cross-device synchronization. The tool operates on a single device (macOS Desktop) and does not persist session information between uses. Each Claude Cowork session is independent. Anthropic has indicated that session persistence and potentially cross-device functionality may come in future updates as the research preview evolves.

How does Claude Cowork handle parallel tasks?

Claude Cowork uses a sub-agent architecture to break complex tasks into smaller components that can be executed in parallel. When beneficial, the main agent spawns sub-agents to handle different aspects of a task simultaneously—for example, processing multiple receipt images at once or analyzing several research documents in parallel. This approach dramatically reduces completion time for multi-file operations while maintaining quality and coherence in final outputs.