Inspiration

Public property and infrastructure information is valuable, but it is fragmented across national registries, map services, municipal portals and local utility sources.

For a homeowner, contractor or adviser, the difficult part is often not knowing the address — it is knowing which authority, dataset or portal to search next.

Forsyningsdata Danmark was created around a simple idea:

The user should know the address, not the public-data bureaucracy behind it.

The WebMCP Challenge made it possible to take that idea further by allowing an AI agent to work with the same application and the same evidence as the person using it.

What it does

Forsyningsdata Danmark is an address-first platform for exploring Danish property, utility and environmental information.

The workflow begins with one address. From there, WebMCP tools can help an agent:

• resolve and verify the address
• retrieve property and parcel context
• retrieve utility information
• perform environmental screening
• identify available map layers
• surface relevant original and local sources
• open the visible address profile
• move the interface directly to the section being discussed
• change interface language while preserving Denmark as the active data country

The important part is that the AI does not replace the website.

The agent can retrieve structured evidence, while the person can still see the same address profile, inspect the information and follow the original sources when an exact drawing, permit, pipe location or authority decision is required.

How it is built

Forsyningsdata Danmark is built with Next.js, React and TypeScript and uses Danish public data sources including Dataforsyningen and other property, planning and geospatial sources.

The application already existed before the challenge as an unfinished public-data product.

During the challenge period, it was meaningfully extended with a WebMCP integration exposing 10 structured tools through the browser's WebMCP interface.

These tools cover interface context, language control, address search, property overview, utility overview, environmental screening, map-layer discovery, follow-up sources and visible interface actions.

The map remains useful as part of the human interface, but WebMCP gives the agent a structured contract instead of forcing it to infer actions from buttons, text and map controls.

Why WebMCP

Property and infrastructure research is a strong example of why WebMCP matters.

Without WebMCP, an AI agent has to interpret a presentation-oriented webpage and guess how to search, navigate and connect information.

With WebMCP, the application can expose explicit actions and structured results while preserving sources, limitations and human-visible context.

For example, an agent can retrieve sewer information and then move the visible address profile directly to the sewer and stormwater section.

The human and the agent are therefore working with the same evidence in the same workflow.

Trust and human control

Not every public-data result has the same level of certainty.

Forsyningsdata Danmark deliberately separates documented information from screening and orientation.

Environmental screening is not presented as a laboratory result. Utility information is not presented as excavation-safe pipe geometry. Missing data is not treated as proof that no risk exists.

When greater certainty is required, the workflow points the user back to the relevant original source.

Challenge and learning

The main challenge was not simply registering browser tools.

It was deciding what each tool should return, where the trust boundaries should be, how structured agent output should relate to visible UI state, and how to keep sources and limitations attached to the result.

WebMCP also introduced a different way of thinking about web applications: not just as interfaces designed for people, but as shared workspaces where people and AI agents can cooperate without hiding the underlying evidence.

That is the direction Forsyningsdata Danmark is now built around.

Built With

• dataforsyning
• geospatialdata
• maplibre
• next.js
• opendata
• typescript
• vercel
• webmcp