Detected: json
Input (JSON / YAML)
Output (TOON)
Output will appear here...
Load Example
Convert JSON and YAML to TOON (Token-Oriented Object Notation) instantly with this free online tool. TOON is a compact data format designed specifically for LLM applications, reducing token usage by 30-60% compared to JSON. All processing happens in your browser — your data never leaves your device.
TOON (Token-Oriented Object Notation) is a data format released in November 2025, designed to minimize token usage when sending structured data to Large Language Models (LLMs). Unlike JSON, which uses quotes, colons, and braces extensively, TOON uses a more compact syntax that LLM tokenizers process more efficiently.
The format was created specifically to address the cost and context-window limitations of LLM APIs, where every token counts. By switching from JSON to TOON for your structured prompts, you can significantly reduce API costs and fit more data within context limits.
Here's how the same data looks in JSON versus TOON:
JSON (verbose)
{
"name": "Alice",
"role": "admin",
"active": true
}TOON (compact)
name: Alice role: admin active: true
Notice how TOON eliminates quotes around keys and string values (when unambiguous), removes curly braces, and uses newlines instead of commas. These small changes compound significantly with larger data structures.
Objects: Key-value pairs separated by newlines, no braces needed for root objects.
name: Alice age: 30
Arrays: Use square brackets with items separated by newlines or commas.
items[3]: apple, banana, cherry
Nested Objects: Inline with curly braces or indented blocks.
user{name: Alice, role: admin}Arrays of Objects: Use indexed notation with field templates.
users[2]{id,name}: 1, Alice, 2, BobWorking with TOON in your code? Here's how to convert JSON to TOON in popular languages:
Typical savings range from 30-60% depending on your data structure. Data with many quoted strings, nested objects, and arrays see the biggest improvements. Use the cost calculator above to see exact savings for your specific data with different LLM providers.
Yes! Modern LLMs like GPT-4, Claude, and Gemini can parse and understand TOON format without any special instructions. The format is human-readable and the models treat it like any structured text. For best results, briefly mention that the data is in TOON format in your prompt.
Yes. All conversion happens entirely in your browser using JavaScript. No data is ever sent to any server. Your input is saved in your browser's localStorage for convenience, but you can clear it anytime.
TOON is optimized for sending data TO LLMs. If you need the LLM to output structured data that you'll parse programmatically, JSON is still better because it has more robust parsing libraries. Also, if your data is already small or you're not hitting context limits, the complexity of adding TOON conversion may not be worth it.
The token counts shown are estimates based on typical LLM tokenizer behavior. Actual token counts vary slightly between models (GPT uses tiktoken, Claude uses its own tokenizer). The estimates are accurate enough for cost comparison purposes — expect ±5% variance from actual API token counts.
Yes. TOON supports all JSON data types: strings, numbers, booleans, null, arrays, and nested objects. The conversion is lossless — you can convert JSON to TOON and back without losing any data.
More DevDen tools: JSON Formatter · Base64 Encoder · URL Encoder · Unix Timestamp · Regex Tester · Hash Generator