The proliferation of online scams and fraudulent communications poses a significant threat to individuals and organizations worldwide. This paper presents an Agentic Honey-Pot system designed to autonomously engage with scammers while extracting actionable intelligence. The system employs a multi-layered approach combining rule-based scam detection with LLM-powered persona agents that simulate believable victim profiles. An enhanced distributed architecture features a central control tower that aggregates data from multiple honeypots to perform cross-attacker profiling and adaptive fraud classification. Built with FastAPI, SQLite, and Groq LLM. Version 1.0.
Samnit Mehandiratta (Mon,) studied this question.