David Golverdingen — Insights

An MCP server is a standard way to give an AI application access to outside systems.

Ask where our vans are around Utrecht, and the model answers with a map.

Enterprise AI reference architectures have converged on two pictures.

Early this year I was wiring Claude into Jira through an MCP server.

I thought the deliverable was the file format and the board.

If an agent writes most of the code, what stops it from confidently shipping the wrong thing?

Enterprise AI gets sold as something only large companies can afford.

Stop building around the model.

The observations in this post came from an eleven-server, 90+ tool production snapshot across eleven external APIs.

Writing good tool descriptions is the start.

An academic analysis of 856 tools across 103 MCP servers (MCP Tool Descriptions Are Smelly!

Most MCP servers do the same thing: wrap an API, expose tools with one-sentence descriptions, and hope the model figures out the rest.

Enterprise AI investment tripled to $37 billion in 2025.

Most MCP writing in 2026 falls into two buckets: marketing for a vendor's product, or tutorials for a hackathon demo.