MCP Crash Course
This is a demo of an MCP server following this YouTube tutorial, MCP Agentic AI Crash Course With Python
To run
uv run mcp dev server/weather.py to run weather mcp
uv run server/client.py to run the client agent which calls the MCP
uv run mcp install server/weather.py to install mcp server locally to Claude desktop
Add the server manually to VS code
.vscode/mcp.jsonjson
{
"servers": {
"my-weather-mcp-server-c23722d0": {
"type": "stdio",
"command": "uv",
"args": [
"run",
"--with",
"mcp[cli]",
"mcp",
"run",
"<absolute-path>/mcp-crash-course/server/weather.py"
]
}
},
"inputs": []
}
Containerized
The mcpserver directory contains another server and client. The Dockerfile containerizes the server which needs to run before calling the client. client-stdio.py doesnt work. Dockerfile-clientsse containerizes the client-sse but doesn't allow the user to input the state code when it runs.
Documentation
-
