This paper provides a systematic regulatory mapping and compliance architecture for AI agents operating under the complex web of EU laws, concluding that high-risk agents with untraceable behavioral drift cannot currently meet the AI Act's essential requirements.