The Model Context Protocol (MCP) is transforming the way AI agents interact with enterprise systems and APIs. While traditional APIs are primarily designed for human developers, MCP creates a bridge that integrates AI models as fully-fledged actors within enterprise landscapes. This allows them to access systems directly and become part of the system architecture. In this presentation, we address the central question of how MCP differs from classical APIs, where parallels exist, and what implications this has for design, integration, and governance. Through a hands-on showcase, we demonstrate a concrete implementation: we show how existing APIs can be made accessible via MCP servers and present a comprehensive MCP blueprint. Additionally, we explore how Azure cloud services can be leveraged to deploy, manage, and secure an MCP server. The talk provides practical insights from the development and operational deployment of MCP servers, including new challenges, solution approaches, and lessons learned.

‍

In economically challenging times, the pressure to make digital systems more efficient continues to grow. At the same time, complexity and the pace of change are increasing - making technical missteps more costly than ever. Yet optimization efforts often focus primarily on the early phases of a project, while lifecycle costs, adaptability, and long-term sustainability are not always given the attention they deserve. At the same time, significant potential can emerge - or quietly be lost - during ongoing operations. The key insight is this: those who prioritize maintainability, resource efficiency, and continuous improvement from the start not only reduce costs, but also build systems that remain affordable, reliable, and effective in the future. With this session, we aim to provide practical food for thought, share real-world experiences, and spark a broader discussion about building IT systems that are both economically and sustainably sound.

‍

The Model Context Protocol (MCP) is transforming the way AI agents interact with enterprise systems and APIs. While traditional APIs are primarily designed for human developers, MCP creates a bridge that integrates AI models as fully-fledged actors within enterprise landscapes. This allows them to access systems directly and become part of the system architecture. In this presentation, we address the central question of how MCP differs from classical APIs, where parallels exist, and what implications this has for design, integration, and governance. Through a hands-on showcase, we demonstrate a concrete implementation: we show how existing APIs can be made accessible via MCP servers and present a comprehensive MCP blueprint. Additionally, we explore how Azure cloud services can be leveraged to deploy, manage, and secure an MCP server. The talk provides practical insights from the development and operational deployment of MCP servers, including new challenges, solution approaches, and lessons learned.

‍

Cloud platforms face a fundamental challenge: How do we enable developers to access cloud and Kubernetes resources quickly and autonomously - without compromising security, compliance, and governance? The Cloud Operating Model describes how responsibilities are distributed between platform and application teams. In doing so, it reveals an apparent contradiction: maximum autonomy on the one hand, centralized control on the other. This session demonstrates how Crossplane resolves this tension at a technical level. As a Kubernetes-native control plane, Crossplane brings governance and self-service together, making the Cloud Operating Model practically achievable. Platform teams define reusable, secure infrastructure patterns, while development teams consume them through self-service APIs - without being able to bypass established standards. Through declarative blueprints, automated validation, and role-based access controls, governance evolves from a bottleneck into an enabler of speed and security. Using concrete real-world examples, the talk illustrates how organizational requirements can be translated into technically enforceable patterns—and how teams can move to the cloud faster, more securely, and more efficiently.

‍