In practical work with AI agents, a question quickly arises: how can they access our internal knowledge bases and tools in a controlled manner, rather than through random prompting? One solution to this problem is MCP servers.

In this session, we will show what MCP is and how to use it to build your own set of tools for AI agents. We will walk through a practical scenario: building our own MCP server, deploying it, connecting it, and integrating it with a developer tool so that the agent can use data and actions in a more structured way.

We will also hold a quick showdown: an agent with MCP versus an agent without MCP. We'll check which one reaches the goal faster, which provides more sensible answers, and which is less prone to creative hallucination.

If you want to see what a basic MCP server architecture looks like, how to expose your own tools and data to an agent, and in which cases MCP genuinely simplifies work versus when it might be an unnecessary addition - this session is for you.