Speakers

We started with a successful B2B API, but quickly realized that feeding it directly to an LLM serving human travelers required a translation layer.  This talk explores how we built a consumer-facing MCP server that bridges that gap — translating rigid API schemas into AI-friendly tools that understand "budgets" and "vibes," deploying a single codebase across multiple AI partner integrations with different auth models and platform constraints, and ensuring reliability through an LLM evaluation framework with "golden prompts," data minimization, and prompt injection defenses.  Read More...