Key Points
- Anthropic released Claude Cowork as a research preview for its Claude chatbot.
- The feature is currently available only in the macOS app for Claude Max subscribers.
- Claude can access a user‑granted folder to read, edit, or create files.
- Typical use cases include reorganizing downloads, generating spreadsheets, and drafting reports.
- Integration with services like Asana, Notion, PayPal, and Chrome is supported.
- Users can queue multiple tasks for parallel execution, reducing back‑and‑forth interaction.
- Anthropic highlighted safety risks such as accidental file deletion and prompt‑injection attacks.
- The feature builds on the earlier “Skills for Claude” initiative introduced in October.
- Non‑subscribers can join a waitlist to gain future access.
Anthropic Expands Claude with Cowork Feature
Anthropic announced a new capability for its Claude chatbot called Claude Cowork. Marketed as a research preview, the feature is currently limited to the macOS version of Claude and is accessible only to subscribers of the power‑user tier, Claude Max.
Claude Cowork enables a user to grant the AI access to a folder on their computer. Once access is provided, Claude can read, edit, or create files within that folder. The company illustrated the functionality with examples such as reorganizing a user’s downloads by sorting and renaming each file, creating a new spreadsheet that lists expenses from a collection of screenshots, and producing a first draft of a report from scattered notes.
The AI agent offers regular status updates on its progress, allowing users to monitor ongoing work. In addition, Claude Cowork can be linked to external services through existing connectors, including Asana, Notion, PayPal, and other supported partners, as well as to Claude in Chrome for browser‑related tasks.
Anthropic emphasized that the new workflow reduces the need for users to repeatedly provide context or reformat Claude’s outputs. Users can also queue multiple tasks, letting Claude handle them in parallel, which the company describes as feeling less like a back‑and‑forth conversation and more like leaving messages for a coworker.
Background and Safety Considerations
Claude Cowork builds on Anthropic’s earlier “Skills for Claude” initiative announced in October, which allowed Claude to load custom instructions, scripts, and resources from folders to improve performance on specific work tasks such as Excel operations or adhering to brand guidelines. Users could also create their own Skills tailored to their jobs.
Alongside the launch, Anthropic warned about potential risks. Because Claude can perform real‑world actions on a user’s device, unclear instructions could lead to destructive outcomes, including the deletion of local files. The company also highlighted the threat of prompt‑injection attacks, where malicious text embedded in referenced content could cause the model to bypass safeguards and perform harmful actions, such as exposing personal data.
Anthropic noted that agent safety—securing Claude’s real‑world actions—remains an active area of development across the industry.
Availability and Access
Claude Max subscribers can try the feature by clicking “Cowork” in the sidebar of the macOS app. Users who are not subscribed can join a waitlist to gain access later.
Source: theverge.com