JSON Schema Tools
Describing an agent’s tools using JSON Schema, the standard way to spell out what valid inputs look like.
Think of it like
A blank tax form with strict rules — this box is a number, that one’s a date, this one’s required.
Example
A tool is defined with a JSON Schema saying its “amount” must be a number and its “currency” must be one of USD, EUR, or GBP.
How it actually works
JSON Schema is a widely-adopted standard for describing the shape of JSON data — types, required fields, enums, nesting — and it became the lingua franca for defining LLM tools. Providers use it to constrain and sometimes to hard-enforce the model’s output, so a call either matches the schema or is rejected. The upside is interoperability; the friction is that expressive schemas get verbose, and every token spent describing tools is context you don’t get back.
For product teams
Standardizing on JSON Schema means your tool definitions work across models and frameworks instead of being bespoke.
For engineers
Tool inputs are specified as JSON Schema, letting the runtime validate or constrain the model’s emitted arguments.
Related
- Function Schema — The general concept it implements.
- Function Calling — What it makes reliable.
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