# Building Your First MCP Server and Client (Self-Paced) Canonical URL: ## Overview This self-paced workshop brings you face to face with the Model Context Protocol (MCP) by guiding you through one complete build from start to finish. You'll come to understand what MCP is, the reasons it matters, and how to craft MCP servers aimed at AI agents rather than passing them off as flimsy covers over REST APIs. Leaning on Python and FastMCP, you'll assemble a live MCP server on top of the SpaceX API, put it through local testing, hook it up to a host that supports MCP, and then write a pared-down MCP client capable of locating and calling the tools your server exposes. The course gives implementation and design an even footing. Past simply standing up a server and client that run, you'll discover how to keep tool sprawl in check, organize everything around the way agents operate, rein in context-window costs, and turn out tool descriptions and outputs that sit well with large language models. By the end, you'll have a working MCP server and client to your name, a grounded grasp of MCP architecture and how tools get invoked, and a repeatable approach you can reuse to design servers that feel far more agent-native. ## What you'll learn - Lay out what MCP is for and walk through how the host, client, and server connect. - Separate tools, resources, and prompts from one another, and pin down when each one fits. - Explain the way context windows shape the design of an MCP server. - Spot the usual missteps in MCP server design, above all the habit of wrapping each REST endpoint one for one. - Stand up a local MCP server in Python with FastMCP. - Shape tools around outcomes and workflows instead of the structure of the API. - Put an MCP server through its paces using the MCP Inspector and an MCP-compatible host. - Assemble a lean MCP client that links to a server, pulls up its tools, and runs them. - Outline what goes into deploying a remote MCP server and the security concerns that come with it. ## Pricing **Tuition:** $649