Tool Calling: How LLMs Use Your APIs Without Running Code

With tool calling, we give the LLM a menu of tools, each with a name, a purpose and its inputs. The LLM asks for one ("call lookup_employee for Rahul"), our code runs it, and the result goes back to the LLM.

The restaurant analogy

• You (the LLM) order from the menu.
• The waiter (our code) takes the order to the kitchen (the tool).
• The waiter brings back the food (the result).
• You never enter the kitchen.

A simple example

User: "What's the weather in Pune?"

LLM  → requests get_weather("Pune")
App  → calls the weather API → {"temp": 28, "sky": "sunny"}
LLM  → "It's 28°C and sunny in Pune."

The LLM never runs code itself. It outputs a structured tool request, my application runs it and returns the result.

- The short version

Next in the GenAI series: MCP (Model Context Protocol): The USB-C Cable for AI

Start from the top: GenAI Quick Answers and Series Guide