In my talk I gave a demo on setting up a very simple MCP server that just says hello if the AI tool detects that the user has entered a greeting.
The key insight
This was something it took a while for me to get my head around — specifically that the interpretation happens before the MCP is called and then the usage and result that comes back from the MCP is deterministic. It’s just code. Which we’re familiar with.
The AI part is in deciding which tool to call and what parameters to pass. The tool itself is a normal function that does a normal thing. Once you understand that split, MCPs become much less intimidating.
What I’d do differently
In hindsight, I wish I had spent more time on explaining how the actual Power BI MCP works. The hello world example is good for understanding the concept, but the real value is in seeing how a practical MCP connects to a semantic model, constructs a DAX query, and returns results.
Part of my MCP Servers for Conversational BI talk.