AI company Anthropic has introduced a new tool, Cowork, and its development story is drawing attention across the tech world. The company says Cowork was written almost completely by its own AI system, Claude Code.
Boris Cherny, the creator of Claude Code at Anthropic, in an X post revealed the details. He wrote, “All of it”, when asked how much of Cowork had been authored by the AI tool.
What is Cowork?
According to TechCrunch, Anthropic launched Cowork on Monday. It is described not only as an easier-to-use alternative to the Claude Code but is, in fact, integrated directly within the Claude Desktop application.
Also Read | What is Angelic Intelligence? 5 key points about this brewing term we bet you didn't know
Cowork lets you choose a specific folder to read from or write to with Claude. You simply use a straightforward messaging interface, making it unnecessary to use command-line interfaces and complicated virtual environments. Anthropic explains it’s like a sandbox Claude Code with a design geared toward individuals who aren't tech-savvy.
Why Anthropic built Cowork
Anthropic says Cowork was inspired by how users were already pushing Claude Code beyond software development. Many were using it as a general-purpose, agent-like AI for everyday tasks rather than purely coding work.
In a post, reminder, Cherny noted that users had applied Claude Code to activities such as planning trips, creating slide decks, organising email inboxes, cancelling subscriptions, recovering wedding photos, monitoring plant growth, and even managing household devices.
How the tool works
Cowork is powered by the Claude Agent SDK and has shared core capabilities with Claude Code. Its permissioning by folder provides more clarity to the user regarding what files they can either read or edit, making it more welcoming to non-technical individuals, according to Anthropic.
Also Read | How artificial intelligence solved the mystery of Italian surgeon Dr Ivaldo’s disappearance
They mention some of its applications, such as receiving pictures of receipts for automatic expense report generation, managing multimedia libraries, scanning social media posts, or analysing communication records.
The company suggests potential use cases, including preparing expense reports from receipt photos, organising media files, scanning social media content, and analysing conversation logs.
Like Claude Code, Cowork can complete a set of tasks without requiring continuous interaction between the model and the user. Anthropic has warned of potential dangers, such as prompt injection or unwarned file deletion. However, these issues relate to dangerous potential use rather than posing a risk to the model's stability or functionality.
