MCP and property management: why owners should start demanding it now

Conrad Wahlén is CTO and co-founder of Vyer. We asked him six questions about MCP — what it is, why it matters and what property owners should do now.

MCP — what is it?

Simply put, MCP, Model Context Protocol, is to AI what an open API is to integrations. It is a standardized way to give AI models and AI agents access to tools, data and functions in other systems.

Think of it this way: instead of asking Claude, ChatGPT or Copilot to draft an email that you then copy into Outlook, with MCP you can ask the AI to create a draft with the recipient, subject and content already filled in.

AI goes from generating a well-worded generic response to being able to act in your systems. That is the difference.

Why should property owners start demanding MCP support from their vendors now?

AI is only as useful as the tools and data it can access. MCP is a way to give AI access to more tools and more of the data that already exists in the organisation.

The earlier those connections are in place, the faster you can develop and test new processes that involve AI.

Building an MCP interface can take time for many systems, especially if there have been no open APIs before. Vendors who start early will therefore have a head start over those who treat MCP as a future concern.

That is why it is right to start asking the question now.

What is the concrete value? What can an organisation actually do with MCP that it cannot do today?

It is about lowering the threshold and making greater use of the digital data that already exists in the organisation.

Take a concrete example: a property manager wants to know which ventilation units in the portfolio are more than 15 years old and have had more than three work orders in the past year. Today that requires manual compilation from several systems. With MCP, an AI agent can retrieve installation data from Vyer, combine it with the work order history and deliver the answer and an analysis from a single question.

It is not new data. It is the same data, available in a completely different way.

What happens to property owners who have their data locked in systems without open connections?

They become limited to what their vendors decide to build and allow. If the function, view or report you need does not exist, you wait for the system to implement it.

That is the same dynamic that existed before open APIs became standard. The organisations that pushed for open interfaces early were able to get more value from their data faster.

The same thing will happen with MCP.

If you are a property owner, what should you do now?

Start by asking your system vendors how they think about open APIs, MCP and AI agents. Find out what data is available today and what the limitations are.

It is not just about data. Many of the tools the organisation already has also become more accessible. Perhaps it has been too complicated to generate a report in the finance system, or too time-consuming to compile a large number of data points. With MCP, someone can ask an AI agent for that report and get a result that was previously difficult or unavailable.

You do not need to build an AI agent tomorrow. But you should make sure that the data and tools you want to use with AI are actually accessible.

The organisations that build that foundation today will have far more opportunities to use AI tomorrow.

Many IT departments see security risks in opening up systems to AI models. What do you say to them?

They are right to ask the question. But many of the risks already exist when systems are exposed through open APIs.

What is required is similar to what good API security already demands: clear permission boundaries, rate limiting, logging and the ability to review what has happened. MCP is no exception to those principles.

Avoiding MCP for security reasons does not solve the problem. It just pushes it forward.

What is the most common misconception about MCP?

That the threshold is high and it is complicated to get started.

It does not have to be. If your data is available digitally and you start with a defined use case, the threshold is lower than most people think. You do not have to build everything at once. Start small, learn how it works in practice and build from there.

The ceiling is then very high. With MCP, AI can not only read data but also act in systems, functioning like a user interface for a programme via an agent. But you get there step by step.