Function Calling

Function calling (also called tool use) is the ability of AI models to determine when a user's request requires calling an external function or API, and to generate the structured parameters needed to make that call. It bridges the gap between AI conversation and real-world actions.

How function calling works: the AI model is given descriptions of available functions (name, description, parameters), during conversation the model decides when a function should be called, the model generates structured JSON parameters for the function call, the application executes the function and returns the result, and the model incorporates the result into its response.

For example, if a user asks "What's the weather in New York?", the model recognizes this needs real-time data, generates a function call like getWeather({location: "New York"}), receives the weather data, and responds naturally: "It's currently 72F and sunny in New York."

In AI agents, function calling enables: appointment booking (calling a calendar API), CRM updates (creating leads in HubSpot), data retrieval (looking up order status), payment processing (initiating transactions), email sending (triggering messages), and any other API-powered action.

On platforms like Chipp, function calling is exposed through "Custom Actions" and "Pro Actions (MCP)" — visual builders that let non-developers define what external functions their AI agent can call, when to call them, and what parameters to pass. The AI decides during conversation when to use these tools based on the user's needs.