Anthropic has made Claude Security available in public beta for all Enterprise customers. The tool, formerly known as Claude Code Security, scans codebases for vulnerabilities and suggests targeted patches. This is done using the Claude Opus 4.7 model.
The tool has been available for the past two months in a limited research preview for hundreds of organizations. Their feedback has shaped the current version, which is now available via the sidebar on Claude.ai or directly at claude.ai/security.
In February 2026, Anthropic launched Claude Code Security, a limited preview for Enterprise and Team customers, to scan complex code vulnerabilities. That name has since been changed to Claude Security.
How Claude Security works
Unlike conventional scanners that use pattern recognition, Claude Security approaches code like a security researcher. The model reads the source code, tracks data flows, and analyzes how components interact across files and modules. After a scan, it provides a detailed overview for each finding: how confident the model is that a vulnerability is real, how severe it is, and how it can be reproduced. Additionally, it generates instructions for a targeted patch that can be applied directly in Claude Code on the Web.
Claude Opus 4.7 forms the foundation of Claude Security. That model was released in mid-April as a significant step forward in software engineering and is the first to incorporate the security measures from Project Glasswing. Early users of Claude Security reported that they went from scan to patch in a single session, something that previously required days of coordination between security and engineering teams.
Partners and availability
In addition to direct access, technology partners—including CrowdStrike, Microsoft Security, Palo Alto Networks, SentinelOne, TrendAI, and Wiz—are integrating Opus 4.7 into their own platforms. Firms such as Accenture, BCG, Deloitte, Infosys, and PwC are helping organizations roll out Claude-integrated security solutions for vulnerability management and incident response.
New features in the public beta include scheduled scans, exporting findings as CSV or Markdown, and webhooks to Slack, Jira, and other tools. Administrators can enable Claude Security via the admin console. Access for Team and Max customers will follow shortly.