# Building Your First MCP Server and Client

Canonical URL: <https://www.creativelive.com/classes/building-your-first-mcp-server-and-client>

## Overview

Across this hands-on workshop, you'll get to know the Model Context Protocol (MCP) by completing a full build from beginning to end. The goal is to leave with a clear sense of what MCP does, the problems it solves, and the thinking behind servers that are built for AI agents rather than thrown together as thin wrappers over a REST API. Python and FastMCP give you the foundation to stand up a real MCP server on top of the SpaceX API, run tests on your own machine, plug it into a host that speaks MCP, and write a lean client able to find and trigger the tools your server offers.

Construction is only half of it; the other half is knowing what makes a good design. As your server and client take shape, you'll pick up strategies for keeping the number of tools manageable, shaping everything around the way agents actually operate, holding down context-window expenses, and phrasing tool descriptions and results so that large language models can make sense of them.

## What you'll learn

- Articulate why MCP exists and lay out how the host, client, and server fit together.
- Tell tools, resources, and prompts apart, and judge which one suits a given situation.
- Shape MCP servers around the way agents work, the results they aim for, and the limits of the context window.
- Catch the usual MCP design missteps, such as wrapping each REST endpoint one for one.
- Stand up and test a local MCP server in Python with FastMCP, MCP Inspector, and an MCP-compatible host.
- Put together a lean MCP client that links to a server, retrieves its tools, and runs them.

## Pricing

**Tuition:** $649