David Golverdingen is a AI engineering lead at Warmtebouw, a 350-person Dutch company that installs and maintains heating, cooling and ventilation, with a small IT team and no AI department. He built and runs its production MCP platform: 18 servers and 114 tools over the company's ERP, BIM, building automation and energy data, with eight MCP Apps that render the answers. In the 90 days to the end of October 2026, 128 colleagues used it through general-purpose AI assistants, across project management, engineering and the back office. Most of them do not write code. He also built the IT team's AI-native development workflow; since it went live, the team merges about 250 pull requests a month, up from about 100.

He wrote The Missing Layer, which introduces Introspective Context Engineering for MCP, a method for turning a thin API wrapper into a Rich Domain MCP Server. His open-source mcp-metadata-demo runs the same server as a thin, a rich and a best version, publishes the evals behind each design rule, and ships a skill that runs the loop against your own server. He spoke at AGNTCon + MCPCon Europe and VibeKode, speaks at Update Conference Prague in November and NDC AI Oslo in December, and writes at davidgolverdingen.nl.