AI-driven attack path discovery system leveraging Claude and Cursor LLM agents with AWS MCP Server integration to identify critical vulnerabilities across AWS infrastructure. The architecture connects multiple data sources—AWS Secrets Manager, GitHub repositories, and filesystem endpoints—through Model Context Protocol servers that feed attack graph data into Amazon Neptune for comprehensive threat visualization. Claude Code and Cursor agents query the attack graph via Graph API Gateway while maintaining least-privilege IAM roles and secrets.read permissions to simulate real-world privilege escalation scenarios. This pattern enables security teams to proactively discover multi-step attack chains before adversaries exploit them, combining LLM reasoning with AWS native services for continuous threat intelligence. Fork and customize this diagram on Diagrams.so to model your own AWS account topology and IAM permission boundaries.