Anthropic has released the Claude Security plugin in beta, bringing AI-powered vulnerability scanning directly into Claude Code for developers who want to catch high-severity flaws before they ship.
The tool lets teams scan recent changes or full repositories from the terminal, using the same Claude inference they already run.
The Claude Security plugin works inside Claude Code, Anthropic’s terminal-based AI agent for editing code and handling git tasks. Developers can scan uncommitted changes before a commit or run a full codebase analysis without leaving their workflow.
Admins enable the feature in the admin console, and setup follows a straightforward getting-started guide. The plugin is available in beta for all Claude Code users.
This approach keeps security checks inside the daily development loop rather than forcing context switches to separate scanners.
Claude Security uses a multi-agent system that first maps the code architecture, then identifies potential threats, verifies findings to cut false positives, and finally suggests style-matched patches.
Humans still review and apply every fix the tool never auto-commits changes. Early users praise the smooth workflow and the verification step that builds trust in results. Large full-repo scans can drive up token costs and sometimes shift to more expensive models, so teams may prefer scoped or incremental runs.
The system reasons through code like a security researcher. It reads Git history, traces data flows across files, and understands business logic capabilities that go well beyond simple pattern matching.
Issues Claude Security Plugin Targets
Claude Security focuses on serious vulnerabilities that traditional tools often miss, including:
- Memory corruption
- Injection flaws
- Authentication bypasses
- Complex logic errors
It excels at context-dependent issues that span multiple files, where understanding how components interact is essential. Severity classifications and confidence rankings help security teams prioritize what matters most.
Rule-based tools catch known patterns but generate high false-positive rates and struggle with novel or logic-heavy bugs. Claude Security analyzes code holistically, generating real fixes tailored to the project’s style instead of generic alerts.
Findings can flow into existing workflows via webhooks for Slack, Jira, or other ticketing systems. Teams can also export results as CSV or Markdown for audits. Documented dismissals carry forward, so triage decisions stay visible across tools.
Users can limit scans to a specific directory or configure scheduled runs for continuous coverage without manual starts. This supports both quick pre-commit checks and ongoing repository hygiene. The plugin complements, not replaces, existing security stacks, feeding results into the tools teams already trust.
The Claude Security plugin is live in beta for all Claude Code users. Organization admins turn it on through the admin console at claude.ai/admin-settings/claude-code.
Open-source maintainers and enterprise teams gain an additional layer of defense against the rising tide of sophisticated, AI-assisted exploits.
By embedding researcher-grade analysis inside the terminal, Anthropic aims to raise the security baseline for codebases of every size.
Developers who already rely on Claude Code can start scanning today and keep vulnerabilities out of production with far less friction than legacy tools require.
The Privilege Paths Attackers See That You Don’t: BeyondTrust Pathfinder Platform Does It for You -> Get Free Identity Security Assessment
The post Anthropic Launches Claude Security Plugin to Scan Code for Vulnerabilities appeared first on Cyber Security News.