AI powered WAF

AI WAF (AI-Powered Web Application Firewall) is an intelligent security middleware designed to protect Large Language Model applications and AI agents from sophisticated adversarial attacks. It defends against prompt injection attempts, jailbreaks, data exfiltration, and social engineering attacks that traditional firewalls cannot detect. The system employs a three-layered defense architecture: fast heuristic analysis using 20+ regex patterns for immediate threat detection, multi-provider LLM intent analysis using a "council of judges" approach with models from Anthropic, OpenAI, Google, and AWS Bedrock, and output validation to prevent system prompt leakage and PII exposure. Built with TypeScript and Express, AI WAF provides a RESTful API for easy integration, real-time security dashboard for monitoring, and configurable policies per AI agent. It features smart optimization that reduces analysis costs by 60-80% through intelligent LLM skip logic while maintaining robust protection. The system analyzes both user inputs before they reach AI agents and validates AI responses before they reach users, providing comprehensive bidirectional security. Ideal for customer-facing chatbots, enterprise AI assistants, and any production LLM application requiring defense-in-depth security.