VulnOdin

Modern enterprise software development has accelerated toward continuous integration and deployment (CI/CD), yet penetration testing remains periodic, manual, costly, and slow. Meanwhile, cybersecurity communities have observed a sharp rise in agentic-AI-driven offensive capabilities, where autonomous multi-agent systems can chain reconnaissance, scanning, exploitation, and privilege escalation with minimal human oversight. This project presents VulnOdin, an AI-powered autonomous red teaming system designed specifically for DevSecOps pipelines and enterprise environments. VulnOdin accepts live URLs and complete codebases including Dockerfiles, autonomously builds and deploys the target environment, constructs a dynamic attack graph, orchestrates traditional offensive security tools (e.g., nuclei, sqlmap, Burp extensions), and executes controlled exploitation attempts under strict governance constraints. The system continuously evaluates applications during development, automatically producing evidence-backed, compliance-ready penetration testing reports aligned with OWASP, CVSS, and ISO frameworks. Our findings demonstrate that AI-supported agentic orchestration can improve exploit chain discovery, reduce false positives through tool-verified evidence, and dramatically shorten the vulnerability discovery cycle. This work proposes a new paradigm: reproducible, continuous, environment-aware, autonomous pentesting integrated directly into enterprise build pipelines.