Securing AI agents: How AWS and Cisco AI Defense scale MCP and A2A deployments

This article describes how AWS and Cisco AI Defense partnership secures AI agents through automated scanning of Model Context Protocol (MCP) servers, Agent-to-Agent (A2A) protocols, and Agent Skills.

• AI Registry provides centralized visibility and governance for MCP servers, agents, and Skills
• Automated security scanning detects vulnerabilities before components become operational
• Three scanner types: YARA Analyzer for pattern detection, LLM Analyzer for semantic analysis, Cisco proprietary scanners
• Vulnerable assets automatically disabled with detailed reports requiring admin review
• Integrates with CI/CD workflows, ServiceNow, Slack, SIEM systems for enterprise workflows
• Open architecture supports REST API compatibility with Anthropic's MCP Registry
• Addresses visibility gaps, security bottlenecks, and compliance audit trail requirements

The partnership enables enterprises to scale AI agent deployments securely with automated governance, reducing manual review bottlenecks while maintaining regulatory compliance.