GuardVibe

GuardVibe is a security MCP (Model Context Protocol) tool designed to verify code written by AI assistants. It addresses four structural gaps: flagging vulnerabilities published after a model's training cutoff via daily CVE updates, running deterministically for consistent results, analyzing cross-file context across entire repositories, and providing independent review of AI-generated code.

The tool includes 520 security rules, 39 analysis tools, and operates fully locally with zero configuration. It integrates with Claude Code, Cursor, Gemini CLI, and other AI coding environments. A unique feature called secure_prompt shifts security left by analyzing coding prompts before generation, embedding security requirements into the AI's execution rather than catching issues after code is written.