Tool Use

Tool Use refers to the capability of large language models (LLMs) to invoke external functions, APIs, or services during inference. Rather than generating only text, a model receiving a prompt can emit a structured “tool call” that triggers an external system and incorporates the result into its ongoing reasoning.

The concept was popularized as “Function Calling” by OpenAI with GPT-4 in 2023 and quickly adopted across all major LLM providers. By 2026, Model Context Protocol (MCP) (MCP) emerged as the industry-standard interface for declaring and connecting tools to agents, unifying what had previously been fragmented proprietary implementations. In Agentic AI systems, tool use is the “Act” phase of the ReAct loop; in Multi-Agent Orchestration, it defines the primary responsibility of each executor agent.