The Agentic Control Plane: Measuring Post-Quantum Readiness in MCP and A2A

AI Agents are now more widely used in production deployments, and they are no longer merely prototypes. Model Context Protocol (MCP) and Agent-to-Agent (A2A) are emerging as the default protocols for how agents communicate with tools, data, and each other. ​ This talk analyzes the state of post-quantum readiness in the emerging agent ecosystem and shows at which layers PQC actually matters. We scan a growing list of public MCP and A2A endpoints identified through open registries, GitHub, and well-known endpoint probes to measure internet exposure, authentication signals, TLS posture, mTLS detection, and post-quantum key exchange. For A2A agents, we also analyze whether Agent Cards are signed, which JWS algorithms are used, and what that implies for future signature migration. ​ Our findings show that some endpoints have started to negotiate hybrid PQC key exchange, while the certificate identity layer remains classical. Further, we attempt to distinguish the CDN edge termination from the origin posture because edge-only numbers overstate the readiness. Legacy TLS versions still appear on deployment infrastructure. The key exchange layer matters now because of the harvest-now-decrypt-later risk, while the signature layer matters for agent identity, Agent Card provenance, and long-lived signed metadata. For publicly trusted WebPKI TLS certificates, certificate-layer PQC is not yet an option until the CA/B Forum finalizes PQC profiles and root programs accept them, but agents using Private PKI can begin piloting ML-DSA certificates. ​ By the end of this presentation, the participants will have a practical checklist about how to assess post-quantum readiness of the agent infrastructure, including what to inventory, what to scan, which post-quantum signals are relevant today, certificate-layer changes to prepare for in the future, and finally how to address crypto-agility for your agent infrastructure.