Anthropic launches Cowork, a file management AI agent that could threaten dozens of startups
Anthropic has launched Claude Cowor, a general-purpose AI agent capable of manipulating, reading and analyzing files on a user’s computer, as well as creating new files. The tool is currently available as a “search preview” only to Max subscribers with plans of $100 or $200 per month.
The tool, which the company describes as “Claude Code for the rest of your work,” leverages the capabilities of Anthropic’s popular Claude Code software development wizard, but is designed for non-technical users rather than programmers.
Many have pointed out that Claude Code is already more of a general-purpose agent than a developer-specific tool. It is capable of creating applications that perform functions for users on other software. But non-developers were put off by the name Claude Code and also by the fact that Claude Code must be used with a coding-specific interface.
Some of the use cases presented by Anthropic for Claude Cowork include rearranging uploads, turning screenshots of receipts into expense spreadsheets, and producing first drafts from notes on a user’s desktop. Anthropic described the tool, which can operate autonomously, as “less like walking back and forth and more like leaving messages for a colleague.”
Anthropic reportedly built Cowork in about a week and a half, largely using Claude Code itself, according to Claude Code director Boris Cherny.
“This is a general agent that seems well-positioned to bring the extremely powerful capabilities of Claude Code to a wider audience,” Simon Willison, a UK-based programmer, wrote of the tool. “I would be very surprised if Gemini and OpenAI don’t follow suit with their own offerings in this category.”
Enterprise AI race
With Cowork, Anthropic now competes more directly with tools like Microsoft’s Copilot for the enterprise productivity market. The company’s strategy of starting with a developer-focused agent and then making it available to everyone could give it an advantage, as Cowork will inherit Claude Code’s already proven capabilities rather than being built as a consumer assistant from scratch. This approach could make Anthropic, which has already reportedly overtaken its rival OpenAI in terms of enterprise adoption, an increasingly attractive option for companies looking for AI tools that can manage their work autonomously.
Like any other AI agent, Claude Cowork presents security risks, particularly around “rapid injections,” where attackers trick LLMs into changing course by inserting malicious, hidden instructions into web pages, images, links, or other content found on the open web. Anthropic addressed the issue directly in the announcement, warning users of the risks and offering advice such as limiting access to trusted sites when using the Claude extension in Chrome.
The company acknowledged, however, that the tool was still vulnerable to these attacks, despite Anthropic’s defenses: “We have built sophisticated defenses against rapid injections, but agent security, that is, the task of securing Claude’s actual actions, is still an area of active development in the industry… We recommend that you take precautions, especially while you learn how it works.”
The launch also sparked concerns among startup founders about the competitive threat posed by large AI labs packaging agent capabilities into their core products. Cowork’s ability to handle file organization, document generation, and data extraction overlaps with that of dozens of AI startups that have raised money to solve these specific problems.
For startups building applications on models from larger AI companies, the concern about foundational AI labs building similar functionality as part of their core product is common. In response to these concerns, many startups have argued that companies with deep domain expertise or better user experience for specific workflows can still maintain defensible market positions.
