Heijmans - MCP
AI & Data
Semester programme:Complex Software Systems
Research group:Future Software
Project group members:Antonio Davidov
Nikola Dakov
Bianca Cristea
Miguel Werneck
Oscary Cijntje
Project description
The Heijmans MCP project explores the development of a secure Model Context Protocol (MCP) server that integrates Relatics with Microsoft Copilot. The aim is to centralize project information and reduce manual workflows by enabling users to retrieve and modify Relatics data through AI-assisted interactions.
The project is guided by the following main research question: How can an MCP server be developed securely to bridge Relatics and Microsoft Copilot to centralize project information and reduce manual workflows? To answer this question, two sub-questions are investigated:
- How can MCP be implemented to enable structured retrieval and modification of Relatics project data?
- What security and authentication mechanisms are required for enterprise-grade deployment?
Context
The Heijmans MCP project is situated within the construction and infrastructure domain, where large projects involve extensive documentation, requirements, risks, and stakeholder information managed in Relatics. Project teams often spend significant time manually searching for information and performing repetitive administrative tasks. This project aims to improve efficiency by integrating Relatics with Microsoft Copilot through a secure Model Context Protocol (MCP) server. The solution enables AI-assisted access to project data while ensuring that security, authentication, and authorization requirements are met for use in an enterprise environment.
Results
The most important outcome of the project is a functional MCP server that securely connects Relatics with Microsoft Copilot. The server exposes structured tools that allow users to retrieve and modify Relatics project data through natural language interactions. Implemented functionalities include the retrieval of requirements and risks, updating existing records, and creating new entries within Relatics.
A second key outcome is the implementation of an enterprise-grade security architecture. The solution incorporates OAuth 2.0 authentication with Microsoft Entra ID, role-based authorization, and controlled access to MCP tools. These mechanisms ensure that only authorized users can access or modify project information, which is essential when handling sensitive project data.
The value of these outcomes was validated through technical testing and stakeholder feedback. Functional validation demonstrated that the developed MCP tools correctly interacted with Relatics and returned accurate information. Security validation confirmed that authentication and authorization mechanisms behaved as intended. In addition, demonstrations with stakeholders provided feedback on the usability and relevance of the solution within existing workflows.
From a Technology Readiness Level (TRL) perspective, the project can be positioned at approximately TRL 5–6. The developed prototype has been validated in a relevant environment using real project data and realistic usage scenarios. Although further work is required to achieve full operational deployment within Microsoft Copilot, the project demonstrates the technical feasibility and practical value of MCP as a secure bridge between AI assistants and enterprise information systems.