Function calling

Definition

A model capability that lets the AI output structured tool invocations (API calls, database queries, etc.) rather than plain text. Function calling is what turns a chatbot into an agent that can take real-world actions.

Why it matters

Function calling is the technical capability that unlocks agentic AI. Without it, models can only generate text; with it, they can interact with the world. Function calling enables models to search databases, send emails, update CRMs, trigger workflows, and chain together complex multi-step operations. For product developers, function calling quality, how reliably a model invokes the right function with correct arguments, is often more important than raw text generation quality. A model that writes beautiful prose but calls the wrong API endpoint is useless for automation. This is why function calling benchmarks are increasingly important for model evaluation.

In practice

OpenAI introduced function calling in June 2023 and it has since become standard across all major model providers. Anthropic's Claude supports tool use with JSON schema definitions. Google's Gemini provides similar capabilities. In a typical implementation, you define functions with JSON schemas, the model returns structured function calls instead of text, your code executes the function, and the result is fed back to the model for further processing. MCP (Model Context Protocol) from Anthropic extends this pattern with a standardized protocol for tool discovery and invocation, reducing the custom integration work required for each new tool.