1 Oct2025 NTK25 Recap: Copilot and MCP for Azure DevOps in Action What an incredible experience! The full rooms and engaging hallway conversations made the week fly by. NTK Konferenca is always a party with a purpose, featuring excellent speakers, a friendly crowd, and numerous opportunities to connect, share ideas, and learn from one another. Excellent speakers make the NT Conference thrive! What We Brought We focused on the everyday application of emerging MCP servers. The Azure DevOps MCP server exposes the most essential parts of ADO to GitHub Copilot inside VS Code, and we demonstrated how it keeps developers in flow and lets product owners stay in charge of decisions. Check out the following sessions: AI‑Powered Developer Workflows: Copilot + MCP + Azure DevOps Open‑source ADO MCP, plus our custom lightweight 7pace MCP server that removes the friction of time logging. This turns natural language into precise time‑tracking actions. AI‑Boosted Product Ownership: Work Smarter, Ship Faster — With Your Brain Switched On A human‑in‑the‑loop approach where GitHub Copilot, driven by a custom product‑ownership prompt, crafts a hierarchical product backlog from a specification, and the product owner edits and accepts, steering the product in the desired direction. Key Concepts The slides below offer an insight into our sessions. They demonstrate how MCP integrates Copilot with Azure DevOps and 7pace, showing how to wire it up in VS Code and explaining how this benefits product owners and developers in their day-to-day work. MCP Architecture - One host, multiple servers, and each exposes focused capabilities The host can connect to multiple MCP servers simultaneously. Each server provides a selected set of actions or context (local or remote), keeping the design modular and secure. MCP Flow - How the MCP client and server coordinate tool calls, resources, and prompts. This sets the context: the host (VS Code/Copilot) talks to an MCP server that exposes tools and resources. Requests are routed securely, and responses are returned to the model. Custom 7pace MCP Server Design - Natural language turns into precise time‑tracking API calls to Azure DevOps A lightweight custom server handles everyday 7pace actions (log time in a work item) so developers avoid context switching and keep focus in VS Code. Azure DevOps MCP in VS Code - Azure DevOps exposed to Copilot through MCP, right inside VS Code The ADO MCP server surfaces projects, work items, backlogs, queries, and more to Copilot, allowing it to fetch context or act precisely without guesswork. Where AI Can Help Product Owners - Discovery, delivery, and communication Examples of everyday gains include research scans, slicing work, acceptance criteria, tests, stakeholder notes, and sprint goals, with the PO deciding at each step. What AI Can / Can’t Do - Where AI fits in the workflow — assistant, not decision‑maker AI excels at preparing and cross-referencing information, but final decisions must remain human. The key for the PO is to treat AI as a teammate, while they drive the product vision and make trade‑offs. Creating a Hierarchical Product Backlog in ADO via GitHub Copilot Prompt and Azure DevOps MCP - Human‑in‑the‑loop flow from specification to Epics/Features/PBIs Copilot proposes backlog items based on a specification; the PO reviews and approves them. MCP executes only approved actions, preserving control and traceability. Takeaways MCP provides Copilot with a clear and safe way to act on Azure DevOps. Developers avoid context switching; product owners keep quality checks. You don’t need a big platform project — small, custom MCP servers go a long way. Put It into Practice If you’d like to see this in your development environment, book a short demo or invite us to run an AI-powered product ownership workshop. We’ll tailor it to your setup and keep it practical.