MCP Is a Protocol. Your Architecture Still Matters.
The Model Context Protocol standardizes how agents talk to tools. It does not decide what a good tool is, who owns it, or what it is allowed to do. Those are still architecture decisions.
The Agentic Architect with Hugo Guerrero
Notes for the people building the capabilities, context, controls, and operating models behind useful AI agents.
The Model Context Protocol standardizes how agents talk to tools. It does not decide what a good tool is, who owns it, or what it is allowed to do. Those are still architecture decisions.
My takeaways from KubeCon + CloudNativeCon North America 2026: agents strain the pod model, sandboxes compete for the runtime layer, identity becomes the control point, and what Solo.io shared in Salt Lake City.
We spent two decades making APIs easy for developers to discover and consume. Agents change the consumer, and the old lessons apply in a stricter form.
Proactive agents need integrations, live events, and compute that stays reachable. A better sandbox on a Mac mini is not production. The real requirement is an agent that is always available, not always running.
Ask an agent for the weather and ten messages cross five components. Agentic AI is not only an AI problem. It is a distributed-systems problem, and it needs distributed-systems discipline.
A reading guide to The Agentic Architect: six essays on the APIs, events, context, memory, and control beneath useful, trustworthy AI agents, in the order that makes the argument easiest to follow.
Agent quality is bounded by the architecture around the model. Each theme covers one part of it.
The contracts and capabilities agents inherit: API design, discoverability, and outside-in architecture.
3 articlesHow context is assembled, moved, and governed: MCP, tools, retrieval, and the paths data takes.
3 articlesThe connectivity layer around agents: gateways, registries, servers, and the controls that govern them.
5 articlesMaking agent behavior dependable: deterministic boundaries, state, history, and evidence.
1 articleReal-time context for agents: events as products, streams as memory, and contracts for change.
1 articleStart here
Generative AI shifts enterprise software from predefined workflows to intent-driven experiences—and makes architecture the foundation of UX.
Read the essay