If context is king, architecture is the castle

The future of AI agents isn't just about smarter models—it's about architecture. In a candid conversation at the AI Agent Conference, Apollo GraphQL CEO Matt DeBurglis argues that context is the new king, but only if it's anchored in a principled, battle-tested architecture. He warns that while AI agents will democratize software creation, unleashing a wave of ad hoc tools and workflows, the real risk isn't technical failure—it's security collapse. With agents acting like high-privilege microservices, the 'East-West' attack surface is exploding, and traditional north-south security won't cut it. DeBurglis insists that organizations must treat this as a system design problem, not a tooling one. He champions GraphQL and the MCP (Model Context Protocol) standard as foundational layers that provide introspection, discovery, and governance—enabling agents to search, execute, and reason safely. The key insight? The most powerful organizations won't be those with the most advanced models, but those with the most resilient architecture. The episode reveals a critical tension: speed vs. safety. While companies are racing to deploy agents, DeBurglis cautions against sacrificing data integrity for velocity. He calls out the 'fox and hen house' problem—trusting models to secure themselves—and urges platform teams to draw clear lines between deterministic, auditable systems (like GraphQL) and the unpredictable, non-deterministic agent layer.