What is Tool Use?
Also known as: Function Calling
Tool Use (aka Function Calling) lets AI agents call external tools, APIs, or functions by outputting structured requests instead of just text.
The model is given a list of available tools with their names, parameters, and descriptions. During generation it can emit a special structured output (often JSON) that specifies which tool to invoke and what arguments to pass.
After the tool executes, its result is fed back into the model’s context so it can continue reasoning or produce a final answer. This loop turns a language model into an agent that can fetch data, run code, or control software.
Tool use is usually combined with prompting strategies such as ReAct or chain-of-thought to decide when and how to call tools.
Example
A user asks an assistant for today’s weather in Paris. The model recognizes it needs live data, outputs a function call to a weather API with the argument “Paris”, receives the temperature, and then replies in natural language.
Why it matters
Tool use extends LLMs beyond static knowledge, enabling them to interact with real-world systems and perform practical tasks, which is essential for building reliable AI agents.
Frequently asked questions
No. Tool use relies on prompting and structured output formats; the model’s weights are not changed.