Secure AI agent access patterns to AWS resources using Model Context Protocol

This AWS guidance outlines deployment patterns for AI agents and MCP servers and presents three security principles—assume all granted permissions could be used, provide organizational role governance, and differentiate AI-driven from human-initiated actions—along with practical IAM policy examples, session-tagging/context-key differentiation, monitoring recommendations (CloudTrail/CloudWatch), and governance controls (permission boundaries, SCPs, session policies) to securely manage agent access to AWS resources.