Anthropic has updated Claude Security so Enterprise customers can scan repositories as their code changes and track findings until a fix lands. The public beta also adds project knowledge, editable threat models and connections to development tools. Teams can investigate problems across repositories, while access controls determine who can see findings or manage projects.
Anthropic has updated Claude Security, its service for finding vulnerabilities in software, with continuous scanning and automatic fix tracking. Enterprise customers can use the new features in public beta. Teams connect code repositories, select the branch and scope, and begin with a deep baseline scan. Code changes are examined according to their risk and complexity.[1], [2]
The findings list follows the code after that first scan. A patch can be written by a developer or another tool; once it lands, Claude Security marks the associated issue as fixed. If the same vulnerability appears again, the finding can reopen. Projects may include several repositories, allowing checks across connected services. Teams can edit generated threat models and supply deployment details and existing controls to inform later scans.[1]
A new application programming interface lets teams manage projects and findings from other software. Its access tokens carry the permissions of the user who created them. The model context protocol connector links findings to Claude and Claude Code, where an engineer can start work on a fix. Administrators can separately authorize viewing, triage and project creation. Code changes still enter the team’s development workflow for assessment.[1]
