Anthropic Launches Claude Cowork Feature for MacOS Users
Key Points
- Cowork lets Claude access a MacOS folder for file organization and document creation.
- Available only to Claude Max subscribers at $100 per month.
- Supports connectors for app integration and works with the Claude Chrome extension.
- Anthropic advises using Cowork on non‑sensitive data during its research preview phase.
- The company has built defenses against prompt‑injection attacks but warns of potential risks.
Anthropic introduced Cowork, a new capability for its Claude AI that lets subscribers grant the chatbot access to a MacOS folder. Users can chat with Claude to organize files, rename items, and generate spreadsheets or documents from the folder's contents. The feature, currently limited to Claude Max subscribers at $100 per month, also links to connectors for app integration and works with the Claude Chrome extension. Anthropic cautions that Cowork is in a research preview, recommending use only on non‑sensitive data and noting defenses against prompt‑injection attacks.
Overview
Anthropic announced a new feature called Cowork for its Claude AI chatbot. Cowork allows users to give Claude access to a folder on their MacOS computer. Once access is granted, users can converse with Claude to have the AI organize files, rename items to follow a naming convention, and read the files to create spreadsheets or other documents. The feature is currently available only to MacOS users with a Claude Max subscription priced at $100 per month.
Key Capabilities
Cowork works by letting users select a folder for Claude to access. In practice, a user could point Claude at a folder of receipts, and the AI would organize the files and produce a spreadsheet listing all expenses. Anthropic also added the ability for Cowork to link to connectors, enabling Claude to interact with other apps to create documents, presentations, or other projects. Additionally, Cowork integrates with the Claude Chrome extension, allowing the AI to perform tasks that require browser access.
Security and Limitations
Anthropic stresses that Claude cannot edit anything it has not been given access to. Because Cowork is still in a research preview—a beta‑style testing period—the company warns that the AI could cause unintended consequences. Users are advised to employ Cowork only on non‑sensitive items. Anthropic also highlighted the risk of prompt‑injection attacks, where an AI might inadvertently execute commands from malicious web content. The company says it has built defenses against such attacks, but acknowledges that a successful attack could let an attacker control Claude, which in turn could act on the folder containing sensitive information.
Industry Context
Cowork arrives as more AI models gain “agentic” capabilities, meaning they can perform actions beyond simple chat, such as booking travel, compiling reports, or ordering food. Similar tools like ChatGPT Agent and Gemini Agent already fulfill user tasks. Google’s Gemini 2.5 Computer Use model also enables AI to interact directly with a computer. The trend extends to retail, where AI standards are being developed for cross‑retailer shopping and payment integration.