An open-source LLM guardrail tool powered by Jev System One that analyzes text submissions across user input, retrieved content, tool output, and model output contexts to provide allow/review/block decisions with ten OWASP risk assessments. The package includes a local Node.js demo, supports deployment to Vercel, and applies semantic scoring with versioned decision policies based on the OWASP LLM Top 10 2026 taxonomy.