Background
Statistics Denmark has thousands of datasets, and CVR contains information about Danish companies. Their APIs require complex calls, and the returned data can consume many unnecessary tokens when an AI system uses it directly.
My approach
I built MCP servers (Model Context Protocol) that act as intelligent bridges between AI models and Danish data sources. The servers transform and compress data so AI models can work with them efficiently.
DST MCP: Statistics Denmark
- Access to all of Statistics Denmark’s database tables
- Smart transformation reducing tokens by 80-95%
- Metadata caching for fast lookups
CVR MCP: Central Business Register
- Search and look up Danish companies
- 95%+ test coverage
- Structured output optimized for LLM consumption
Results
Both servers are public open source. They demonstrate how MCP can give AI systems useful access to complex data sources while reducing token consumption in the measured calls.