Function calling (tool use) lets AI models invoke external APIs, query databases, and perform actions in the real world. These 227 models support structured tool definitions - essential for building AI agents and automated workflows.
You provide the model with a list of available functions (tools) and their parameter schemas. The model decides when and how to call them based on the conversation context, returning structured JSON with the function name and arguments.
Function calling is the foundation of agentic AI. Models can chain multiple tool calls to complete complex tasks - searching databases, calling APIs, modifying files, and reporting results in a single conversation turn.
Combined with JSON mode, function calling ensures the model produces machine-readable output. This is critical for integrating AI into production systems where reliability and parseable responses are required.
Customer support bots that look up orders, coding assistants that run tests, research agents that search the web, data analysis pipelines that query databases - all powered by function calling.
Function calling (tool use) allows AI models to invoke external functions, APIs, or tools. The model decides when to call a function, formats the arguments, and processes the results — enabling AI agents and automated workflows.
Most major models support function calling: GPT-4o, Claude 3.5, Gemini 2.0, and many open-source models. Support quality varies — some models are better at choosing the right function and formatting arguments correctly.
Yes, "function calling" and "tool use" are different names for the same capability. OpenAI calls it "function calling," while Anthropic uses "tool use." Both describe the model's ability to invoke external tools.