Tool Use: Definition and Examples

Tool Use refers to the mechanism by which a large language model (LLM) can call external functions or services during a conversation. Instead of only producing text, the model identifies that a specific action is needed, formulates a structured call (usually in JSON) to the appropriate tool, and then integrates the result into its response.

Concretely, the developer defines a set of available tools with their parameters (name, description, JSON schema). When the user asks a question, the model decides whether to use a tool, which one to choose, and what arguments to pass. The application then executes the call and returns the result to the model, which interprets it to formulate a natural response.

This approach solves several fundamental limitations of LLMs: the lack of real-time data, the inability to perform reliable calculations, and the impossibility of acting on external systems. Thanks to Tool Use, an assistant can check the weather, book a flight, query a database, or execute code — all while maintaining a fluid conversation.

Tool Use is at the heart of the transition from chatbots to AI agents. It is what enables a model to go from "I can talk about it" to "I can do it". Major providers (Anthropic, OpenAI, Google) all offer native implementations of this feature, with variations in syntax but an identical principle.