Technology

Tool calls

Tool calls enable Large Language Models (LLMs) to execute external functions, APIs, or code, extending their capability beyond text generation to perform real-world actions.

Tool calling—often termed function calling—is the mechanism that transforms an LLM from a passive text generator into a proactive, agentic system. It works by having the model intelligently output a structured JSON object (not executing the code itself) that specifies which external tool to call and with what parameters (e.g., calling a `get_weather` function with `location: Paris`). This capability allows the AI to access real-time data, execute complex calculations, automate workflows, and interact directly with third-party systems like CRMs or databases, fundamentally bridging the gap between conversational AI and operational software.