A developer built a basic currency converter to understand the Model Context Protocol (MCP). The example shows how AI can now directly use external tools like a get_exchange_rate function instead of just describing where to find data. The architecture involves an API that lists available tools to Claude AI, which then decides when to call them.

This simple project demonstrates tool discovery and parameter extraction. The AI parses natural language like "¥10,000 for a $60 game" and automatically structures the call. The key takeaway is that MCP lets developers create focused, simple tools while the AI handles the complex connection to human intent.

The pattern scales beyond currency. Developers are using MCP for crypto trading, database queries, and even running Kali Linux commands via natural language. The protocol's power lies not in complexity but in standardizing how AI discovers and uses external functions. A working demo and full code are available for those wanting to experiment.