ComplyAIgent

ComplyAIgent transforms enterprise compliance from a manual, point-in-time scramble into an always-on, automated certainty. It ingests internal policies and external regulations, classifies every rule into enforceable buckets (code-level scannable patterns, actionable developer checks, infrastructure metadata, and semantic guidance), and blocks violations before they reach production. A three-tier Comparator deduplicates regulatory text using entity anchors, MinHash LSH, and semantic LLM analysis. The Categorizer generates machine-readable guardrails with regex patterns and verification questions. The pg CLI fetches these rules, scans staged git diffs with Gitleaks, regex, and entropy scoring, and blocks high-risk pushes. Medium-risk findings route to human-in-the-loop approval via LangGraph interrupts. Built in 5 days for the AMD Developer Hackathon 2026, ComplyAIgent demonstrates how agentic AI can shrink audits from weeks to hours and give compliance officers, developers, and CISOs a single source of truth — from regulation to enforcement, fully logged and always auditable.