Designing APIs & MCP for AI Readiness

APIs power modern digital systems. But they were designed for human developers building deterministic applications. Today, those APIs are increasingly consumed by AI-driven and agentic systems that are probabilistic, non-deterministic, and autonomous by nature. This mismatch introduces real risks such as runaway costs, unexpected behavior, security gaps, and brittle integrations.

Making APIs “AI-ready” requires more than adding an OpenAPI description or exposing an endpoint to an LLM. It demands rethinking APIs across multiple dimensions: how they are designed, how they are discovered, and how they are governed at runtime. In this session, we explore why traditional API assumptions break down in AI-driven environments and what architectural changes are required to address them. We’ll examine design patterns for AI-friendly APIs, challenges introduced by agentic consumption, and the critical role of API gateways, policies, and observability in controlling risk.