Practical MCP with FastMCP & LangChain

Setting Up A MCP Environment For Learning

For the sake of this guide we'll be looking at different concepts, practicing with code examples, and building applications. To make this more concrete and productive, you will certainly need to use some MCP servers that will provide you with the necessary knowledge to understand the protocol and how to use it in practice. I recommend using the following tools:

• GitHub Copilot (on VS Code) as your AI pair programmer.
• Claude Code as your code assistant.
• Claude Desktop as your AI assistant.

There are more than these 3 tools that support MCP, you can choose any other tool that you like.

When in doubt, you can ask any of these tools to help you understand MCP concepts, write code examples, or debug your applications.

Let's now detail how to integrate my recommended MCP servers into some tools.

First of all, download the following file from FastMCP website: https://gofastmcp.com/llms-full.txt:

Example:

curl -o $HOME/docs/llm.txt https://gofastmcp.com/llms-full.txt

The file llms-full.txt contains the whole documentation of FastMCP in a format that can be easily parsed and consumed by AI tools.

Next, install npm if you don't have it already.

VS Code

Whenever you create a new workspace or open a project, add the following file to .vscode/mcp.json:

{
        "servers": {
            "deepwiki": {
                    "url": "https://mcp.deepwiki.com/mcp",
                    "type": "http"
            },
            "github": {
                    "url": "https://api.githubcopilot.com/mcp/",
                    "type": "http"
            },
            "fastmcp": {
                "command": "npx",
                "args": [
                    "-y",
                    "@modelcontextprotocol/server-filesystem",
                    ""
                ],
                "type": "stdio"
            },
            "langchain": {
                "url": "https://docs.langchain.com/mcp",
                "type": "http"
            }
        }
}

Now launch these servers using the Start button that appears ontop of each server. Once started, you can reference them in your discussions with GitHub Copilot. For example:

#fastmcp Explain how to create a tool in FastMCP and how to call it from a client.

#deepwiki What is the difference between a tool and a resource in MCP?

#github How do I use function calling in Langchain?

#langchain How Langchain can help me build an MCP server?

#fastmcp #langchain How to integrate FastMCP into a Langchain agent?

Here is a quick explanation of what each MCP server does:

• deepwiki: A comprehensive knowledge base for MCP. It provides up-to-date documentation for every repo in the world. You can use it to ask questions about the projects and frameworks we are using in this guide.