Dataverse MCP will break most of the Business Apps' logic currently set up in Power Platform

🚨Rant time🚨 Dataverse MCP will break most of the Business Apps' logic currently set up in Power Platform. Let’s cut to the chase: Dataverse MCP servers, coupled with the ongoing AI revolution, are the second wave of god-mode chaos for low-code solutions built on Canvas and Model-Driven Apps. This is not a hyperbole - it's a real challenge that every enterprise Power Platform developer should worry about. Why? Because MCP servers will bring your nightmares to reality as they open up pathways for users (or agents) to interact directly with your data models, bypassing all the carefully designed logic and validation you’ve painstakingly built on the front end. # Nightmare scenario Imagine this: What happens when a user, with a tool such as Level Up for Dynamics (a.k.a. god-mode), bypasses your Model-Driven App JavaScript customizations and interacts with everything from your forms? Answer: total chaos. Now, imagine MCP servers or AI agents doing the same thing. But it turns out it's even worse. They are unpredictable and can integrate directly with your data models, completely disregarding the UI you’ve built to control how data flows. The validations, field restrictions, or even business processes you’ve enforced on the frontend? Gone. ### Why is it an issue for everything? Low-code platforms have always been a double-edged sword. I often say that low code means the never-ending overcoming of roadblocks that platform creators put in front of you. This is a reality - many things are often simpler, which makes them faster to implement, but some things are harder or even impossible. I feel like this is the case, especially for database implementation - let's look into both SharePoint and Dataverse. Have you ever seen a project in which there is a proper RLS implemented on SharePoint lists? I saw a few, but in most cases, the process is to share edit access with everyone and filter data on the frontend. In some cases, there may be a separation between edit and view permissions, but that's the most you can expect. Have you ever wondered or tested what would happen if someone suddenly went through all Canvas apps, scraped SharePoint list references, and removed all data they could access? MCP would be able to do it quite easily. One would say that SharePoint is not an enterprise-grade solution for building apps. Dataverse is more robust - that's for sure. You have model-driven apps that are built specifically for your DATA MODEL, but you always have plenty of logic built on the frontend as well, and that's the issue. AI can interact only with what it sees - your frontend will not matter. This means that all process validation processes or logic that hides specific fields or adds some requirement permissions will be disregarded. Have you ever thought about what will happen in your process if somebody adds the data only into 2 out of 3 tables or submits a change that should not be possible in a given process stage? What if somebody changes the status of the approval process when that column is controlled strictly by your backend process? What if somebody triggers a step 5 steps ahead of the current state and starts sending automated messages to a VP of your organization? That's an MCP nightmare. # What comes next? Bottom line: This isn’t just a breaking change - it’s a workflow wrecking ball for existing apps. MCP, like any other tool, AI can redefine how somebody accesses your data, which makes simplifications of low-code platforms an issue. What's the worst part? It cannot be addressed for all of the issues, as it's just a tool that exposes to you what's always been there. What does it mean? - For Microsoft, it creates a need to allow for the creation of stronger server-side validation tooling that can better interact with the frontend of low-code tools, as without it, Power Platform developers will always opt into going to simple JavaScript manipulations. - For all developers, it means security should be first. Always. Even in low-code solutions, security can no longer be an afterthought. As tools evolve to make data more accessible, you need to proactively ensure that users or agents can’t abuse those capabilities. Application being accessible only by your organization does not mean that it's safe. Treat it as if it's exposed to external threats. For a more serious process, audit your data access and validation workflows with the assumption that someone can bypass your UI. ___