ThreatHunter

ThreatHunter is an AI-powered cybersecurity threat intelligence platform built for Track 1: AI Agents & Agentic Workflows. It helps security teams move beyond raw vulnerability lists by turning package, source-code, and configuration findings into prioritized, evidence-backed remediation decisions. Unlike a RAG chatbot, ThreatHunter does not rely on document retrieval as its core architecture. It uses a tool-driven, auditable multi-agent workflow that coordinates an Orchestrator, Security Guard, Scout, Intel Fusion, Analyst, Critic, and Advisor. These agents gather and validate evidence from OSV, NVD, GHSA, OTX, EPSS, CISA KEV, exploit intelligence, and ATT&CK-style signals. ThreatHunter analyzes exploitability, chained risk, repeated findings, and remediation priority, then produces structured JSON reports with actionable recommendations. Memory, JSONL checkpoints, and a FastAPI + SSE dashboard make the workflow replayable, observable, and suitable for live demonstration. For AMD alignment, ThreatHunter is designed to use an AMD Developer Cloud vLLM/OpenAI-compatible endpoint as the primary inference path, with fallback providers used only for demo resilience.