Skip to content

THR978 Theater Tue 17 Nov 15:30 PST

Hardening MCP: Advanced Security and Scale for Enterprise AI Hubs

EY Kalman

Summary

From the recording

Speaker EY Kalman set out a reference pattern for running MCP at enterprise scale. Agents call through a gateway or load balancer that routes, authorises and rate limits. Each request is now self-describing, and every return path is authenticated. The pattern uses Entra agent identities, Azure Front Door with WAF, API Management, content safety prompt shields, private endpoints, Azure Firewall and API Center as the inventory of approved servers. No GA or preview statuses, dates or pricing were given, and the protocol change was described only loosely.

Session description

Microsoft's catalogue description

Exposing Model Context Protocol (MCP) servers to enterprise production introduces massive scalability and security hurdles. In this advanced, demo-first session, we address the architectural realities of deploying the new stateless MCP standard. Drawing from real-world FastTrack deployments, we will dissect advanced multi-tenant routing, stateless load balancing, and prompt-injection mitigation. Learn how to securely scale AI data pipelines without compromising corporate data boundaries.

← All sessions