Introduction to TOON: A JSON-like Format for LLM Prompts
2026-05-03 7 min read
TOON (Template Object Oriented Notation) is a human-friendly format for structuring LLM prompts and responses. Unlike JSON, TOON supports comments, is more lenient with quotes, and reads more naturally. Our JSON to TOON converter helps you translate between the two formats.
TOON Syntax Basics
TOON example
{
// Define a user profile
user: {
id: 123
name: Alice Johnson // no quotes needed
email: alice@example.com // URLs work too
verified: true
tags: [admin, editor, contributor] // commas optional
}
// Nested objects are intuitive
metadata: {
created: 2026-05-03
source: api
}
}TOON vs JSON
| Feature | TOON | JSON |
|---|---|---|
| Comments | ✓ Supported | ✗ Not allowed |
| Quote strings | Optional | Required |
| Trailing commas | ✓ Allowed | ✗ Invalid |
| Human-readable | Optimized for it | Secondary concern |
| Parser support | Specialized | Universal |
| Best for | LLM prompts, config | APIs, data transfer |
Real-World Use Case: LLM Prompt Structure
TOON prompt template
{
system_prompt: "You are a helpful JSON validator."
// Example inputs for the LLM
examples: [
{
input: {valid_json: [1, 2, 3]}
output: {error: null, message: "Valid"}
}
{
input: {invalid: {trailing: comma,}}
output: {error: "Unexpected token", line: 1}
}
]
// Instructions
instructions:
"Validate each JSON input and return structured feedback"
}Why TOON for LLM Workflows
- Comments make prompts self-documenting
- Lenient syntax reduces parsing errors in LLM-generated output
- More readable than JSON for human prompt engineers
- Smaller than JSON (no quotes, optional commas)
- Native support in emerging LLM frameworks
Convert JSON to TOON
Transform your JSON data into readable TOON format for LLM prompts.