Anthropic brings Claude Security scanning to Claude Code in beta

Anthropic adds security scanning to Claude Code

Anthropic has made its Claude Security plugin for Claude Code available in beta, giving developers a way to scan repositories and code changes for potential vulnerabilities from within a Claude Code session.

According to Anthropic’s Claude Code documentation, the Claude Security plugin is designed as a multi-agent vulnerability scanner that can examine a repository or a diff, validate selected findings, and help turn findings into patches for user review. Anthropic’s Claude Security product page similarly describes the tool as scanning codebases, validating findings, and suggesting patches.

The plugin is part of Anthropic’s broader Claude Code plugin system. Anthropic’s documentation on discovering plugins says users can install prebuilt plugins through marketplaces, including the official claude-plugins-official marketplace, which includes security-related plugins.

What the beta plugin does

Anthropic’s Claude Code documentation says the Claude Security plugin can scan either an entire repository or a diff. That distinction matters for developers who want to evaluate a full codebase as well as teams reviewing narrower changes before they are merged.

The documentation also says the plugin can validate findings. In security tooling, validation is an important step because static analysis and automated review can produce false positives. Anthropic’s product page frames Claude Security as a tool that scans codebases, validates findings, and suggests patches, rather than simply listing possible issues.

Anthropic’s Claude Code documentation further states that selected findings can be turned into patches for user review. That wording indicates the plugin is not presented as an autonomous patching system; instead, developers are expected to review proposed changes before applying them.

Japanese technology publication gihyo.jp reported that Anthropic announced the beta availability of the Claude Security plugin for Claude Code on July 22, 2026. The report says the plugin can scan codebases for vulnerabilities from inside a Claude Code session.

Installation through Anthropic’s plugin marketplace

Anthropic’s plugin marketplace documentation explains that prebuilt plugins can be discovered and installed through marketplaces. For official plugins, Anthropic points to the claude-plugins-official marketplace.

That documentation is relevant because the Claude Security plugin is distributed as a Claude Code plugin rather than as a standalone security product alone. Developers using Claude Code can install plugins from a marketplace and then use them in their coding environment, according to Anthropic’s plugin documentation.

Anthropic’s Claude Security product page says the Claude Security Plugin for Claude Code is now in beta for scanning, validating, and patching inside Claude Code. The beta label suggests the product may still change, and teams evaluating it should treat its results as input for review rather than as a replacement for established security processes.

Why it matters for developers

The launch places security review closer to everyday coding activity. Instead of requiring developers to leave their editor or command-line workflow to run a separate review process, Anthropic is positioning Claude Security as part of Claude Code.

The most concrete capabilities described by Anthropic are repository scanning, diff scanning, validation of findings, and patch suggestions. Those features align with common developer needs: finding security issues in existing code, checking new changes, reducing noise from unverified findings, and preparing fixes for human review.

The sources do not establish how Claude Security performs compared with traditional static application security testing tools, dependency scanners, or manual penetration testing. Anthropic’s documentation and product page describe the intended workflow and capabilities, but they do not provide independent benchmark results in the cited material.

For now, the clearest takeaway is that Anthropic has extended Claude Code with a beta security plugin that brings vulnerability scanning and patch suggestion workflows into its coding assistant environment. Developers can evaluate it through Anthropic’s official plugin marketplace and should review any generated findings or patches before relying on them in production code.