Ask a model to check today’s weather and, on its own, it can’t - it can only generate text. Function calling, also called tool use, is what closes that gap: the model recognizes a task needs something outside itself, like live data or a calculation, and asks for a specific tool or function to run, then uses the result to shape its answer.
Example
Ask an AI assistant without function calling what the weather is in Mumbai right now, and it can only guess or admit it doesn’t know. One with function calling recognizes it needs current data, calls a weather API function, and answers using the result.
n8n relevance
n8n’s AI Agent node runs on function calling under the hood: you give the agent access to specific tools - a Sheets lookup, an HTTP Request, another sub-workflow - and the model decides mid-conversation whether and when to call each one, based on what the task actually needs. That’s what lets an AI agent take real action instead of just generating text.
Related concepts
See also AI agent, MCP, and orchestration.