Function Calling

Function calling (also called tool calling) is a capability where AI models can recognize when a user's request requires external data or actions and generate structured function calls to fulfill that need. Instead of trying to answer from its training data alone, the model outputs a JSON object specifying which function to call and what parameters to pass, allowing the application to execute the function and return results for the model to incorporate into its response.

The function calling workflow operates in a loop. The developer defines available functions with their names, descriptions, and parameter schemas. When a user asks a question, the model decides whether a function call is needed. If so, it generates the call specification. The application executes the function and returns results. The model then uses those results to formulate its final response. This enables AI to access real-time data, perform calculations, interact with databases, and take actions in external systems.

Function calling is the foundation of AI agents and tool-using AI systems. It bridges the gap between language understanding and real-world action. For example, a travel assistant can call flight search APIs, a financial advisor can call stock price APIs, and a coding assistant can call code execution environments. The quality of function calling depends on clear function descriptions and parameter schemas that help the model understand when and how to use each tool.