AI company Anthropic has unveiled a new product called Cowork, and in a striking twist, the tool was built almost entirely by AI itself.
Boris Cherny, the creator of Claude Code at Anthropic, revealed that Cowork was written pretty much fully by Claude Code. When asked on X how much of the product had been authored by the AI tool, Cherny replied bluntly, “All of it.”
(Also Read: Mangaluru man builds AI-powered device that orders food when his stomach growls)
About Cowork?
According to a report by TechCrunch, Anthropic announced the launch of Cowork on Monday. The new tool is designed as a more accessible, user-friendly version of Claude Code and is built directly into the Claude Desktop app. The tool allows users to assign a specific folder that Claude can read from or modify, with instructions provided through a standard chat interface. The setup mirrors a sandboxed instance of Claude Code, but without the technical hurdles like command-line tools or virtual environments.
The company says the product was inspired by how users were already stretching Claude Code beyond programming tasks, using it as a general-purpose, agentic AI system.
In an X post, Cherny said the launch reflects how Claude Code users have already been applying the tool to everyday tasks, ranging from vacation planning and slide deck creation to cleaning up email inboxes, cancelling subscriptions, recovering wedding photos, monitoring plant growth, and even controlling household appliances.
Calling Cowork “early and raw,” Cherny described it as the first step towards making Claude Code useful for all non-coding work. He highlighted features such as a built-in virtual machine for isolation, browser automation support, compatibility with Claude data connectors, and safeguards that prompt users for clarification when instructions are unclear. “We are excited to see how you all use it,” he wrote.
(Also Read: Techie quits ₹19 LPA Bengaluru software job, says AI will replace coders)
How does the new tool work?
Cowork is built on the Claude Agent SDK and uses the same underlying model as Claude Code. Its folder-based permissions system gives users clearer control over file access, making it less intimidating for non-technical users. Anthropic says this design opens the door to a wide range of use cases, such as assembling expense reports from receipt photos, managing media files, scanning social media posts, or analysing conversation logs.
Like Claude Code, Cowork can execute sequences of actions without constant user input - a capability that comes with risks. In its announcement blog post, Anthropic warned users about potential issues such as prompt injection or accidental file deletion, urging them to give clear instructions. “These risks aren’t new with Cowork but it might be the first time you’re using a more advanced tool that moves beyond a simple conversation,” the company said.