Frequently Asked Questions

TOON (Token-Oriented Object Notation) is a compact, human-readable data format designed specifically for LLM applications. It eliminates the redundant syntax in JSON — repeated keys in uniform arrays, excessive brackets, and whitespace — while preserving the full data structure. The result is 30–60% fewer tokens for equivalent data.

Savings depend on your data shape. Uniform arrays of objects (like database query results or API response lists) typically yield 40–60% savings. Deeply nested structures save 20–35%. Simple key-value pairs save 10–20%. Paste your own data into the converter to get an exact count.

Yes. TOON is plain text — any LLM that can parse structured text can work with it. You add a short description of the format to your system prompt, and the model reads it correctly. The converter generates this description on request.

100% of processing runs in your browser using JavaScript. Nothing is ever sent to a server, stored in a database, or logged. There are no tracking cookies and no analytics on your data. You can even disconnect from the internet after the page loads and the converter will keep working.

Yes. TOON conversion is lossless and bidirectional. Paste TOON into the output pane and click Reverse — you get back identical JSON. This makes it safe to use in production pipelines: convert before sending to the LLM, reconstruct after.

The converter safely handles files up to 50 MB. For larger files, split them into batches — most real-world use cases involve files well under 10 MB.

Yes. After the initial page load the app runs entirely in your browser. No internet connection is required for conversion, token counting, or any other feature.

Use JSON for standard REST APIs, database storage, config files for non-AI tooling, and anywhere the consuming system expects JSON. Use TOON when you are passing structured data to an LLM and token count matters — prompts, context windows, fine-tuning datasets.