Anthropic wants you to use Claude to ‘Cowork’ in latest AI agent push

Anthropic wants to expand the capabilities of Claude’s AI agent and take advantage of the growing hype around Claude Code – and it’s doing so with a brand new feature released Monday, dubbed “Claude Cowork.”

“Cowork can support many of the same tasks as Claude Code, but in a more accessible form for non-coding tasks,” Anthropic wrote in a blog post. The company is releasing it 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 through Claude’s macOS app, and only to subscribers of Anthropic’s advanced user tier, Claude Max, which costs between $100 and $200 per month depending on usage.

Here’s how Claude Cowork works: A user gives Claude access to a folder on their computer, allowing the chatbot to read, modify or create files. (Examples Anthropic gives include the ability to “reorganize your uploads by sorting and renaming each file, create a new spreadsheet with a list of expenses from a stack of screenshots, or produce a first draft of a report from your scattered notes.”) Claude will provide regular updates on what he’s working on, and users can also use existing connectors to link it to external information (like Asana, Notion, PayPal and other supported partners) or link it to Claude in Chrome for browser-related tasks.

“You do not need to continue to manually provide context or convert Claude’s results into the correct format,” Anthropic wrote. “You also don’t need to wait for Claude to finish before offering further ideas or feedback: you can queue tasks and let Claude run them in parallel. It feels a lot less like back-and-forth than leaving messages for a colleague.”

The new feature is part of Anthropic’s (and its competitors’) attempt to provide the most useful AI agents for both consumers and businesses. AI agents have come a long way from their humble beginnings as mostly theoretical tools, but there’s still a lot of development to do before you see your non-tech friends using them to accomplish everyday tasks.

Anthropic’s “Skills for Claude,” announced in October, was a partial precursor to Cowork. Starting in October, Claude could get better at custom tasks and jobs, thanks to “folders containing instructions, scripts and resources that Claude can load when needed to make him smarter at specific work tasks – like working with Excel.” [to] following your organization’s brand guidelines,” according to a statement at the time. People could also develop their own skills for Claude in relation to their specific jobs and tasks they were required to complete.

As part of the announcement, Anthropic warned of the potential dangers of using Cowork and other AI agent tools, namely that if instructions are not clear, Claude has the ability to delete local files and take other “potentially destructive actions” – and that with rapid injection attacks there are a range of potential security issues. Rapid injection attacks often involve bad actors hiding malicious text on a website that the model references, which instructs the model to bypass its protections and do something harmful, like transmit personal data. “Agent security, that is, the task of securing Claude’s actual actions, is still an area of ​​active development in the industry,” Anthropic wrote.

Claude Max subscribers test the new functionality by clicking on “Cowork” in the sidebar of the macOS application. Other users can join the waiting list.