Project is forked from zahere-dev/mcp-labs
This project demonstrates how to build a simple MCP (Model Context Protocol) client and server using ollama with llama3.2 model Python SDK and the mcp package. We expose a BMI calculator tool via the server and use an LLM-powered client to discover and invoke the tool based on natural language queries.
Watch the full original tutorial (does not contain the changes related to Ollama and streamable HTTP): How to Build an OpenAI MCP Client and Server
(https://www.youtube.com/watch?v=hMHYhRcd_Uo)
- ✅ Lightweight MCP Server using
FastMCP - ✅ BMI calculation exposed as a tool
- ✅ [Not used anymore] OpenAI GPT-4o model used to intelligently select tools and arguments
- ✅ Ollama with llama3.2 model used to intelligently select tools and arguments
- ✅ STDIO-based client-server communication
- ✅ streamable-http based client-server communication
- ✅ JSON-based tool calling flow using natural language