Claude Code security scans your code for vulnerabilities
Anthropic just dropped Claude Code Security, a new AI tool that scans code for vulnerabilities and even suggests fixes.
Right now, it's in research preview for Enterprise and Team users, with maintainers of open-source projects able to apply for expedited or free early access.
The launch was announced on February 20, 2026.
It can spot tricky bugs that often slip past traditional scanners
Instead of just looking for known patterns like older tools, Claude Code Security uses human-like reasoning to spot tricky bugs—think logic errors or other complex, context-dependent vulnerabilities—that often slip past traditional scanners.
It double-checks its own work to keep false alarms low.
It rates the severity of each issue
Running on the latest Claude Opus 4.6 model, this tool has already uncovered over 500 hidden bugs during real-world security challenges.
Plus, it rates the severity of each issue so teams can see what matters most right from their dashboard.
It helps keep your projects safer with less hassle
Unlike rule-based scanners, Claude Code Security reads and reasons about code context and adapts on the fly—so it catches more subtle issues and helps keep your projects safer with less hassle.