Function Schema
The formal description of a function’s name, inputs, and types that tells the model how to call it correctly.
Think of it like
A form with labeled, required fields — the model can only fill in what the form asks for, in the right format.
Example
The schema says get_weather takes a city string and an optional units field, so the model produces exactly those arguments and nothing else.
How it actually works
A schema turns a vague “call the weather tool” into a strict contract — required fields, types, allowed values — that the model’s output must satisfy. Good schemas with clear descriptions dramatically cut malformed calls; sloppy ones invite the model to hallucinate arguments. It’s the interface layer where natural language meets structured code, and its quality quietly determines how reliable your tool use is.
For product teams
Well-written schemas are the difference between tools that just work and tools the agent constantly misuses.
For engineers
A typed signature (often JSON Schema) constraining the arguments the model must emit for a tool call.
Related
- Function Calling — What the schema enables.
- Tool Schema — A near-synonym at the tool level.
- JSON Schema Tools — The format schemas are usually written in.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome