Claude Security Explained: Anthropic's AI Code Scanner

Last Updated: November 9, 2026

Summary: Claude Security is Anthropic's AI vulnerability scanner. It connects to your code repositories, builds a threat model, scans continuously as code changes, verifies each finding and proposes a patch that a person must review. The managed product is in public beta for Claude Enterprise only, billed at token cost with no platform fee, while a separate Claude Security plugin brings on-demand scans to any paid Claude Code plan.

Searches for "Claude Security" have climbed from near zero in early 2026 to about 880 a month in the US, according to SE Ranking, as Anthropic moved the product from a limited preview in February to a public beta in April and added continuous scanning in October. Most pages that rank are single-product help docs or articles about securing Claude Code itself, which is a different topic.

This guide is for the business or security leader deciding whether Claude Security fits their team. It explains what the product does, who can get it, what it costs, how it compares with Anthropic's other security tools, and the limits to check before you sign up. All details were checked against Anthropic's own pages on November 9, 2026.

Claude Security is a product from Anthropic that uses Claude models to find security vulnerabilities in a company's source code, check that each one is real, and suggest a fix. Anthropic's Claude Security product page says it reasons through code "like a skilled security researcher" instead of matching known patterns.

According to the product page, Claude Security traces data flows across files and repositories, can read Git history, and focuses on issues such as memory corruption, injection flaws, authentication bypasses and complex logic errors, especially context-dependent problems that span several files.

Claude Security is reached from the Claude.ai sidebar or at claude.ai/security, according to Anthropic's April 30, 2026 beta announcement, and needs no API integration or custom agent to start. Admins switch it on in the admin console.

The name causes some confusion in search results. "Claude Security" is the vulnerability scanner. "Claude Code security" usually refers to how safe it is to let the Claude Code agent run on your machine, which is a separate question covered in our Claude Code best practices guide.

Claude Security works in four steps: you connect repositories to a project, Claude writes a threat model and runs a deep baseline scan, it keeps scanning as new commits land, and every finding comes with a recommended patch for a person to approve. Anthropic's help center article says the baseline scan "may take several minutes or hours."

Fix tracking closes the loop. Claude Security checks open findings against changed code and marks them fixed when the fix lands, whether the fix was written in Claude Code or anywhere else. Anthropic adds: "If Claude Security finds the issue again later, the finding reopens automatically."

Each finding carries a severity rating judged against the threat model and an explanation of how an attacker could reach the code, with an optional attack-path diagram. The April announcement lists confidence ratings, likely impact and reproduction steps as well.

The managed Claude Security product is available only to Claude Enterprise customers during its public beta. Anthropic's help center lists three requirements: "An active Claude Enterprise plan, a premium seat for each user, and usage credits enabled for consumption billing."

The April 30 announcement said "Access for Claude Team and Max customers is coming soon," but as of November 9, 2026, Anthropic's help center and product page still describe the beta as Enterprise only. Pro, Max and Team users who want to scan code today have the Claude Security plugin for Claude Code instead, covered below.

Inside an Enterprise organization, an owner turns the product on under Organization settings, and admins choose who can use it. According to the help center, custom roles control who creates projects, and each project has three roles: Full (manage findings, settings and access), Triage (view, update status and start fixes) and View.

Claude Security is billed at direct token cost with no extra platform fee, on top of the Enterprise plan and premium seats it requires. The help center says plainly: "There is no additional platform fee." Token usage covers scans, threat model creation, project chat and fix checks.

Anthropic does not publish a per-scan or per-repository price, and the product page sends buyers to contact sales. That means the real cost depends on the size of your codebase, how often it changes and which models run the scans.

There is a launch offer. According to Anthropic's help center, eligible Enterprise organizations get up to $25,000 a month in free usage credits for two months, through December 31, 2026. The help center adds that unused credits do not carry over. Admins can also set a separate spend limit for Claude Security and see month-to-date and daily cost in project settings.

For the per-token rates behind consumption billing, see our Claude API pricing breakdown.

The Claude Security plugin is a separate, on-demand version that runs a multi-agent vulnerability scan inside a Claude Code session on your own machine. Anthropic's plugin documentation says a team of Claude agents "maps your architecture, builds a threat model, hunts for vulnerabilities, and independently reviews every finding before writing the report."

Anthropic's product page says the plugin is "now available in beta for all Claude Code users," but the plugin documentation lists narrower prerequisites: a paid plan, Anthropic API access or a supported third-party provider such as Amazon Bedrock, plus Python 3.9 or later. Pro users have to turn on dynamic workflows in /config first. You install it from the official marketplace with /plugin install claude-security@claude-plugins-official, then run /claude-security to scan a whole repository, scan only a branch, pull request or commit, or suggest patches.

Key differences from the managed product, according to the documentation:

The documentation also warns that "scans are nondeterministic: two scans of the same code can surface different findings," and recommends running scans regularly. If you are new to plugins, our guide to Claude Code plugins explains how installation scopes work.

Claude Security is one of six security layers Anthropic now describes, and the right choice depends on whether you want hosted monitoring, an on-demand deep scan or a quick review. The table below combines Anthropic's plugin documentation with the product and help pages.

Anthropic is clear that none of these replace your current tools. The plugin documentation says it "doesn't replace your existing source-code security tools" and should run alongside static analysis, dependency scanning and code review, because its reasoning complements their deterministic checks.

For many mid-size companies, a sensible split is the plugin for engineers who want a deep check before merging, Code Review on pull requests, and the managed product only once the organization is on Enterprise and wants continuous coverage with an audit trail.

Claude Security connects to other systems through an API, webhooks, an MCP connector and exports. According to Anthropic's October update, the API lets teams list and read projects and findings, resolve or reopen findings, and create and manage projects and repositories, with personal access tokens limited to the user's own permissions.

Anthropic also lists technology partners embedding Claude Opus 4.7 for security, including CrowdStrike, Microsoft Security, Palo Alto Networks, SentinelOne, TrendAI and Wiz, plus services partners Accenture, BCG, Deloitte, Infosys and PwC.

The main limits are Enterprise-only access, no Zero Data Retention, a cap of eight repositories per project, and the fact that AI findings and patches can be wrong. The help center states: "Zero Data Retention isn't supported," and Anthropic may retain data where required by law or to address Usage Policy violations.

Other limits from Anthropic's help center and documentation:

Data handling is the item most security and legal teams will ask about first. Read Anthropic's trust center and confirm your retention requirements before connecting production repositories, since the managed product cannot run under Zero Data Retention.

Sameer's take: In my conversations with public-sector organizations that are actively piloting AI, their biggest worry is privacy and data security.

Claude Security has gone from a limited research preview to a continuously scanning Enterprise product in about eight months. The timeline below uses Anthropic's own announcements.

The original February announcement is still online at Anthropic's newsroom, which is why some articles still describe Claude Security as a limited preview. For wider context on AI in security, see our AI cybersecurity statistics.

Is Claude Security free?
No. Claude Security is billed at direct token cost with no extra platform fee, and it requires a Claude Enterprise plan with a premium seat per user. Eligible Enterprise organizations get up to $25,000 a month in free credits for two months, through December 31, 2026.

Can I use Claude Security on a Pro or Max plan?
Not the managed product, which is in public beta for Claude Enterprise only. Pro, Max and Team users can run the Claude Security plugin inside Claude Code, which needs a paid plan, API access or a supported third-party provider.

Does Claude Security fix vulnerabilities automatically?
No. Claude Security proposes a patch for each finding, but Anthropic says every patch requires human review and approval. The Claude Code plugin writes patches to a folder and never applies them automatically.

What is the difference between Claude Security and Claude Code Security?
Claude Code Security was the name of the limited research preview Anthropic launched on February 20, 2026. It was renamed Claude Security when it opened as a public beta for Enterprise customers on April 30, 2026.

Does Claude Security support Zero Data Retention?
No. Anthropic's help center says Zero Data Retention isn't supported for Claude Security, and Anthropic may retain data where required by law or to address Usage Policy violations.

Claude Security is Anthropic's answer to a real gap: finding the context-dependent bugs that pattern-matching scanners miss, then proving them and drafting fixes. The managed product is strongest for Enterprise customers who want continuous coverage, fix tracking and an API for their existing ticketing and vulnerability tools.

Everyone else can start with the Claude Security plugin on a paid Claude Code plan. Before rolling out either one, check three things: whether no Zero Data Retention is acceptable for your code, how you will budget token-based scanning once the launch credits end on December 31, 2026, and who on your team owns reviewing every patch. For a broader look at defensive AI, read our guide to AI for cybersecurity.

⭐ Get more AI Business Weekly in Google: add us as a preferred source on Google.

Written by Sameer Khan with AI assistance, fact-checked against the sources linked above.