Function Calling
Provider-side support for models emitting structured JSON that matches a declared function signature, so runtime can safely parse and dispatch tool invocations. Analogous to MCP tool calling but predates MCP as a proprietary provider capability.
Related terms
- Tool Calling
The general pattern where an LLM emits structured output (typically JSON matching a schema) that a runtime interprets as a call to an external function, API, or service. Also called "function calling". MCP standardizes discovery and invocation across providers.
- Structured Output
Model-side support for producing responses that conform to a declared schema (JSON Schema, Pydantic model, OpenAI's response_format). Reduces brittle string parsing in tool-use pipelines and is a prerequisite for reliable agent execution.
- Model Context Protocol (MCP)
An open protocol authored by Anthropic that standardizes how LLM applications discover, invoke, and stream from external tools, resources, and prompt templates. MCP treats tool discovery as a first-class runtime concern rather than requiring per-integration glue code.
Related on this site
Framework dimensions
Free tools
Whitepapers
Comparisons
Next steps
- Take the assessment— See where you stand
Call this framework and its tools from your own agent via the Model Context Protocol (MCP) server. Works with Claude Desktop, Cursor, Zed, Continue, and the OpenAI Agents SDK.
Where this fits
Function Calling is part of the Agentic AI vocabulary used in the Generative AI Maturity Framework. See the full glossary for the complete set of 149 defined terms, or take the free maturity assessment to see where your organisation stands.