There is a lot of noise around AI agents and frameworks like MCP. But as McKinsey recently said, “it’s not about the agent, it’s about the workflow.”

In this talk, I strip away the hype and look at what is really needed for agentic API consumption to work at scale: predictability, discoverability, and deterministic execution. MCP and similar protocols are useful exposure layers, but they do not solve these deeper integration and reliability challenges.

That is where Arazzo comes in. As the co-author of the specification under the OpenAPI Initiative, I will show how Arazzo defines multi-step workflows in a machine and human readable format, allowing AI systems to move from guessing to guaranteed execution paths.

We will look at real examples, tooling support, and how Arazzo aligns with the last decade of API standards to enable safe, observable, and predictable automation.