Securing AI agent identity with X.509 PKI on AWS

AWS and Keyfactor published a reference architecture for giving AI agents cryptographic identity with X.509 PKI, aimed at the gap left by the Model Context Protocol's lack of a built-in authentication standard.

• The post argues that sharing API keys between components or passing OAuth tokens through agent chains falls short of the zero-trust requirements enterprise deployments demand, since any component could otherwise impersonate an agent, replay intercepted credentials, or invoke tools without authorization.
• The architecture combines four mechanisms: signed request envelopes for proof-of-possession, user-signed delegation assertions that authorize agents without sharing credentials, RFC 8693 audience-scoped tokens that limit exposure from a single token disclosure, and real-time OCSP revocation checking that propagates credential changes across agents.
• It uses Amazon Bedrock for LLM inference through the AWS Strands Agents SDK, with Amazon RDS, Amazon SES and AWS Secrets Manager, while Keyfactor EJBCA Enterprise supplies certificate issuance, OCSP validation, CRL distribution and lifecycle management.
• Keyfactor says its machine identity platform is used by more than 1,500 enterprise customers worldwide.
• The post says the same authentication pattern applies to any MCP-compatible tool server, and complete source code is available in an accompanying GitHub repository.