MCP in Practice: Building, Using, and Securing the Model Context Protocol. Welcome & Introduction

At this event, you will get three different perspectives on how the Model Context Protocol is changing the way we build and run AI-integrated systems. From understanding to applying to securing with experts from CrateDB, MongoDB, and Prescriba.

The Model Context Protocol (MCP) is opening new ways to connect AI with code and data — but it’s still early days.

This meetup brings together three developers working at the forefront of MCP to share what they’ve learned about building, using, and securing MCP servers in real projects.

Agenda for the meetup

• Welcome and short introduction
• First talk by Ivan Sanchéz Valencia, CrateDB (30 minutes + 10 minutes for QA)
• Second talk by Rune Kaagaard, Prescriba (30 minutes + 10 minutes for QA)
• Short break with sandwich and networking (20 min)
• Third talk by Gagik Amaryan, MongoDB (30 minutes + 10 minutes for QA)
• Round-up and goodbye

Is this talk for me?

This event is geared toward developers who already have some understanding of MCP and how it connects AI models with tools and data. If you’re brand new to AI or not working with code, you’re still welcome to join — just note that the talks will dive into technical details.

First Speaker: Ivan Sanchéz Valencia, CrateDB

Title: How It’s Made: The model context protocol (MCP)

Abstract

People often joke that there are more developers building MCP servers than developers using them, and there is some truth in that, in part because it’s very easy to create one. In this talk we’ll take a look under the hood at how MCP works and explore some tricks that the Python SDK uses to make building a server a painless experience.

Bio

Iván Sanchez believes in creating and sharing knowledge for free to further science and human development in a sustainable way. Even if it is just centering divs.

He advocates for Open Science and Open Source, having participated in many related projects in the past.

He loves Software Engineering and enjoys diving deep into the details of things, he has worked in web development, backend, mobile development and data engineering, spanning multiple languages, frameworks and techniques.

He is currently neck-deep in the database world working as a Database Environment Engineer @CrateDB, an open-source distributed SQL database.  

--

Second Speaker: Rune Kaagaard, Prescriba

Title: Claude Code in Practice: Making It Work on Real Projects

Abstract

Claude Code is surprisingly good at handling large codebases, but only if you teach it properly. This talk covers practical lessons from working with Claude Code on a sprawling legacy Python and Javascript project. 

We'll look at how best to give Claude Code access to your stuff via MCP servers, teaching it once instead of every session, and techniques that actually helped when dealing with 5,000,000 lines of code.

Bio

Rune Kaagaard is CTO of Prescriba and builds developer tools in Copenhagen. He's built several MCP servers including MCP Alchemy for SQL databases, and co-created Django Hypergen with Jeppe Tuxen, a liveview framework for server-rendered reactive HTML. Before programming, he played jazz piano. He's been contributing to open source for 16 years and still believes beautiful structures are hiding in the complexity, waiting to be found. 

--

Third Speaker: Gagik Amaryan, MongoDB

Title: Secure MCP Servers in Production: It’s all on you

Abstract

MCPs are meant to be these magical tools that let you integrate the latest and greatest AI models into any service. But when you’re dealing with actual customer data and systems, this sure sounds too good to be true. I will talk about the dangers of MCP servers, both when building and using them — from exposing insecure tools to hallucinating agents and the biggest liability of them all: the humans running them.

Bio

Gagik Amaryan is part of MongoDB ’s Developer Tools team, working on its terminal and AI tools, notably the official MongoDB MCP Server. Outside of work, he also develops tetrify , a local-first, chat-centered productivity app.