X.509 PKI Architecture for AI Agents from AWS and Keyfactor

Amazon Web Services (AWS) and Keyfactor have announced a joint reference architecture to close authentication gaps when AI agents access sensitive areas such as databases and APIs. Developed using the Model Context Protocol (MCP) infrastructure, the framework aims for secure data exchange utilizing X.509 public key infrastructure (PKI) and the Keyfactor EJBCA platform.

As AI agents access databases and APIs in business processes, cryptographic authentication requirements have gained critical importance. Traditionally used methods such as sharing API keys or transmitting OAuth tokens fall short of meeting enterprise-grade zero-trust requirements. Especially since the Model Context Protocol (MCP) standard structure does not inherently include built-in authentication, vulnerabilities such as identity spoofing or unauthorized tool calls among agents can pose risks.

The newly published reference architecture bridges this gap through the integration of Keyfactor EJBCA and AWS services. The solution relies on four main security mechanisms: signed request envelopes, user-signed authorization claims, RFC 8693 target-scoped tokens, and real-time OCSP revocation checks. Supported by Amazon Bedrock and the AWS Strands Agents SDK, all sample code for this architecture can be accessed via GitHub.