Back to Glossary Index
Core ConceptModel Output / Tool Invocation Layer

Tool Calling (Function Calling)

Industry Definition Set • Entity Resolution Path: /glossary/tool-calling

Quick Answer / TL;DR

An LLM capability to select and invoke external functions (tools) with structured arguments, enabling models to perform actions beyond text generation.

Key Takeaways

  • LLM outputs structured function call specifications.
  • Host system executes the actual function.
  • Supported natively by GPT-4, Claude, Gemini, and others.
  • MCP standardizes tool definitions across models.
Definitive Statement: An LLM capability to select and invoke external functions (tools) with structured arguments, enabling models to perform actions beyond text generation.

Technical Context & Protocol Usage

Detailed Explanation
Tool calling allows an LLM to output a structured request to call a function with specific arguments. The LLM does not execute the function itself; it generates a function call specification that the host system (like an MCP client) executes. Modern models like GPT-4, Claude, and Gemini all support tool calling natively. MCP standardizes the tool definition and calling interface across models.

Format & Payload Metadata

Format: Structured JSON function call specification

Latency: No overhead from model side; execution depends on tool

Real-World Implementation Use Case

Claude decides to call an MCP `database_query` tool, generating a structured call with SQL parameters that the MCP client executes.

M
MCPserver.in Engineering

Platform Team

Published: 2026-07-20
Updated: 2026-07-20

Cite This Page

MLA Style:

MCPserver.in Engineering. "Tool Calling (Function Calling)." MCPserver.in Knowledge Hub, 20 July 2026, mcpserver.in/glossary/tool-calling.