Anthropic has unveiled a new agentic AI tool named Cowork. This tool represents the startup's latest push to compete with major rivals like OpenAI and Google. Cowork specifically targets the general consumer market, marking a strategic shift for the company.
What Cowork Does on Your Desktop
Cowork operates within the Claude desktop application. It can autonomously perform various actions on a user's computer. These actions include creating spreadsheets, editing and organizing files, and generating reports from scattered notes. The tool is designed to work independently once given a task.
"Once you've set it a task, Claude will make a plan and steadily complete it, while looping you in on what it's up to," Anthropic explained in a blog post. This functionality gives Claude "much more agency" compared to standard chatbot interactions.
Access and Availability
Cowork is currently available in a research preview phase. Access is limited to Claude Max subscribers. Users can find it through the Claude macOS desktop app, where a 'Cowork' option appears in the sidebar.
When granted access to web browsers like Google Chrome, Cowork can also complete tasks that involve navigating the internet. This expands its capability to handle online research and data gathering.
How Cowork Compares to Claude Code
Cowork is built on the same foundations as Claude Code, Anthropic's popular AI coding tool launched in February 2025. However, Cowork serves as a less advanced, more accessible version.
"Cowork is a simpler way for anyone—not just developers—to work with Claude in the very same way," the company stated. This design choice makes the technology available to non-technical users who might find traditional developer tools intimidating.
The tool simplifies the user experience by eliminating the need for constant manual context provision. Users don't have to wait for Claude to finish before offering additional ideas or feedback. They can queue up tasks and let Claude work through them in parallel.
Connecting to External Data
Using existing connectors, Claude in Cowork can link to external data sources. This capability allows it to create documents, presentations, and other files based on real-time information. The integration makes Cowork a potentially powerful productivity assistant.
Safety and Security Considerations
Anthropic has implemented several safety measures for Cowork. Claude in Cowork can only access folders and connectors explicitly chosen by the user. "Claude can't read or edit anything you don't give it explicit access to," the company assured.
The AI agent will ask for permission before taking any significant actions. This feature allows users to steer or course-correct the tool as needed. However, Anthropic acknowledges that Cowork remains susceptible to prompt injection attacks, similar to other agentic AI tools.
"We've built sophisticated defenses against prompt injections, but agent safety—that is, the task of securing Claude's real-world actions—is still an active area of development in the industry," Anthropic added.
Potential Risks and Industry Context
Security researchers have flagged several agentic AI tools as posing potential security risks. These include AI browser agents like ChatGPT Atlas and Perplexity's Comet, as well as IDE platforms such as Google's Antigravity. These tools could open the door to new types of threats, including direct or indirect prompt injection attacks.
Anthropic specifically warned that Cowork could "take potentially destructive actions (such as deleting local files) if it's instructed to." The company recommends users provide "very clear guidance" to prevent misinterpretation of instructions.
Anthropic's Broader Strategy
Until now, Anthropic has primarily focused on capturing enterprise market share with advanced tools like Claude Code. Claude Code has become one of Anthropic's most successful products to date. The company has also rolled out specialized versions of Claude for specific sectors, including healthcare, financial services, and life sciences.
Cowork represents a deliberate expansion into the general consumer market. This move positions Anthropic more directly against consumer-focused offerings from competitors like OpenAI and Google. The startup, reportedly eyeing a $350 billion valuation, continues to innovate across both enterprise and consumer segments.
The launch of Cowork demonstrates how AI companies are increasingly developing tools that bridge technical and non-technical users. As agentic AI becomes more mainstream, accessibility and safety will remain critical concerns for both developers and end-users.
