What is function calling or tool use in an AI API

What Function Calling Actually Enables

Function calling lets a developer describe specific available actions — checking a database, looking up current weather, calling another internal service — to an AI model, giving it a structured way to request that one of those actions happen as part of responding to a conversation.

Who Actually Executes the Function

It’s important to understand that the model itself doesn’t execute anything directly — when it decides a function call is appropriate, it outputs a structured request specifying which function and what inputs, and the developer’s own code is what actually runs that function and returns a result back to the model.

Why This Is the Foundation of Agent Behavior

This mechanism is the underlying foundation of most AI agents’ ability to take real actions rather than only generate text — an agent ‘using a tool’ is, under the hood, generally this same function-calling pattern: the model requesting an action, and surrounding code carrying it out.

How the Model Decides When to Use It

The model decides when calling a function is appropriate the same fundamental way it generates any other output — predicting, based on the conversation and the available function descriptions, that requesting a specific function is the most likely useful next step, not through separate fixed logic distinct from how it generates text.

Bottom Line

Function calling gives a model a structured way to request a specific action be taken, with the developer’s own code responsible for actually executing it — this is the core mechanism that underlies most AI agents’ ability to interact with the world beyond generating text.

Go deeper

• Can Claude Write and Run Code Directly, or Only Suggest It?
• What Is an AI Agent, and How Is It Different From a Chatbot?
• What’s the Difference Between an AI Chatbot and an AI Agent?