Let’s Make and MCP Server (And Skill)!

MCP servers are fundamentally tool-providers to our Agentic Gen AI tools and provide a robust vendor-agnostic solution to building and sharing custom tooling to the overall Open Source community.

And making one is not as hard as some might think. Today we’ll explore making an MCP server as well as a Skill and hosting them both locally and in GCP. We’ll cover usage patterns and wrap with costs and next steps.

This talk will take us through every step from ideation, to initial hello world creation, through to a full solution buildout. We’ll explore streamable HTTP vs STDIO and implement both NPX and container driven solutions. Lastly, we’ll cover hosting as a serverless function in GCP as well as a self-managed option in a local Kubernetes.

Additionally, as our audience often works in the medical space, we’ll talk about specific considerations with regards to FDA regulated domains.

Speaker Bio

With a background as a Cloud and SRE Architect, I have a passion for process and building out scalable solutions. My career has spanned over 20 years experience, and I’ve had the honor of working at local companies such as Medtronic, Abbott, Thomson Reuters, and CH Robinson with a focus on SRE, DevOps and Cloud Migrations.

Today, as a Senior Staff Cloud Architect at Abbott (SJM) in the Structured Heart division, my focus is on transcatheter aortic valve implantation (TAVI) next generation solutions. I live in Woodbury, MN and enjoy camping, canoeing and fishing with three wonderful daughters.

What to bring

If you want to follow along, code samples will be shared.

If you wish to code along with me, plan to have:

• Google Antigravity
  • download for free for any OS
• GCP (for CloudRun function)
  • this has a free trial. However, the cost if you aren’t in a free trial for the cloud run function is less than $1
• Docker (or Podman)
  • for containerizing MCP Servers

I will cover local hosting in Kubernetes (k3s) and Docker (though Podman will work too).

I use Linux so my demos will be in Linux/Ubuntu.

Code Samples

An earlier version of this presentation was given to Open Source North 2026. However, this presentation will be updated (for instance, I focused heavily on Gemini CLI in the OSN 2026, but as Gemini CLI is deprecated, we’ll look at using Antigravity and other GenAI tools)

Feel free to peruse the earlier code examples on Codeberg here

Register

Please register and see more event specifics on the GDG Rochester Event page