GuardRail

GuardRail is an open-source tool that blocks potentially dangerous commands run by AI coding agents before they execute. It provides 13 core free guards that prevent destructive actions like force pushes, mass deletes without WHERE clauses, recursive deletions on critical paths, and credential leaks. It works by intercepting shell commands triggered by AI agents, blocking harmful operations and logging all actions in an audit trail. GuardRail integrates natively with Claude Code and offers additional Pro features for compliance and advanced protections.