Anthropic wants to expand Claude's AI agent capabilities and capitalize on the excitement around Claude Code. And it's doing just that with a new feature released Monday called “Claude Cowork.”
“Cowork can take on many of the same tasks that Claude Code can handle, but in a more approachable way for non-coding tasks,” Anthropic wrote in a blog post. The company is releasing this as a “research preview” so the team can learn more about how people are using it and continue building accordingly. So far, Cowork is only available via Claude's macOS app and only to subscribers of Claude Max, Anthropic's power user tier. Prices range from $100 to $200 per month depending on usage.
Here's how Claude Cowork works: The user gives Claude access to a folder on his computer so the chatbot can read, edit, or create files. (Examples Anthropic cited included the ability to “reorganize your downloads by sorting and renaming each file, create a new spreadsheet with a list of expenses from a pile of screenshots, and create the first draft of a report from scattered notes.”) Claude provides regular updates on what you're working on, and users can use existing connectors to link to external information (like Asana, Notion, PayPal, and other supported partners), or use Chrome to create a new spreadsheet with a list of expenses. You can also link to Claude. Browser related tasks.
“You don't have to keep manually providing context or converting Claude's output to the appropriate format,” Anthropic wrote. “You also don't have to wait for Claude to finish before providing further ideas or feedback. You can queue tasks and have Claude process them in parallel. It's more like leaving a message for a colleague than an interaction.”
This new feature is part of Anthropic's (and its competitors') commitment to providing both consumers and businesses with the AI agents that are actually the most useful. AI agents have come a long way since their humble beginnings as mostly useful tools in theory, but there is still a lot more development to be done before we see our friends outside the tech world using them to complete their daily tasks.
Anthropic's Skills for Claude, announced in October, was a partial precursor to Cowork. Starting in October, Claude can now improve his personalized tasks and work with “folders containing instructions, scripts, and resources that Claude can load whenever he needs to make certain work tasks smarter, such as working in Excel.” [to] People were also able to build their own skills for Claude related to specific jobs and tasks that needed to be completed.
As part of the announcement, Anthropic warned of the potential dangers of using Cowork and other AI agent tools, including the fact that Claude has the ability to delete local files and perform other “potentially destructive actions” if instructions are not clear, and prompt injection attacks, which pose a variety of potential safety concerns. Prompt injection attacks often involve a malicious attacker hiding malicious text on a website referenced by a model. This instructs the model to bypass security measures and perform harmful actions such as handing over personal data. “Agent safety, the task of protecting Claude's real-world behavior, remains an area of active development in the industry,” Anthropic wrote.
If you're a Claude Max subscriber, try the new feature by clicking “Cowork” in the sidebar of your macOS app. Other users can also join the waitlist.
