Tool Calling
The ability of an LLM to reliably output structured data to execute external functions.
Definition
Tool calling (also known as function calling) is the capability of a large language model to recognize when a user request requires an external action and to produce a structured output, typically JSON, that specifies which function to invoke and with what arguments. The model itself does not execute the function; instead, the application layer receives the structured call, runs the function, and optionally feeds the result back to the model for further reasoning.
Key characteristics of tool calling include:
-
Structured Output Generation: The model returns well-formed, schema-compliant data rather than free-form text, enabling reliable programmatic consumption by application code.
-
Schema-Driven Definitions: Developers define available tools with names, descriptions, and parameter schemas. The model uses these definitions to decide when and how to invoke each tool.
-
Multi-Tool Orchestration: Modern models can select from dozens of tools in a single conversation and chain multiple tool calls sequentially or in parallel to complete complex tasks.
-
Foundation for Agents: Tool calling is the primitive that enables agentic behavior. Without reliable function invocation, agents cannot interact with external systems, databases, or APIs.
Tool calling is supported by all major model providers including OpenAI, Anthropic, Google, and open-source models through standardized interfaces.