en.Wedoany.com Reported - U.S. AI company Anthropic has released Claude Security as an official plugin for Claude Code, allowing paid-tier users to launch multi-agent codebase scans directly in local sessions. The plugin studies project architecture, builds threat models, and assigns vulnerability search tasks to multiple agents, with each finding independently verified before being added to the final report.

Claude Security can analyze the entire code repository, individual components, branch changes, pull requests, or specific commits. The generated report includes exploitation scenarios, severity levels, model confidence, and remediation suggestions. For selected vulnerabilities, the plugin can generate patches, which are created in a separate repository copy and additionally verified by another agent. The source code is not automatically modified; developers must manually apply the patches.

Users can install the plugin from the Anthropic official marketplace using the command "/plugin install claude-security@claude-plugins-official". After installation, scanning is initiated with the command "/claude-security". The runtime environment must meet the requirements of Claude Code 2.1.154 or higher, Python 3.9.6+, and Dynamic Workflows support. The scanning process consumes the current plan's quota.

Previously, this tool was released under the name Claude Code Security for a limited research preview. Additionally, a standalone cloud version, Claude Security, is used to monitor connected code repositories and is currently available as a public beta for Enterprise customers. Documentation: https://code.claude.com/docs/en/claude-security