FAQ
Is my data uploaded anywhere?
No. Parsing, conversion and token counting all run in your browser — the site works even offline once loaded. There is no backend and no upload.
How much does TOON actually save?
Typically 30–60% of tokens versus formatted JSON for uniform arrays of objects. For nested or irregular data the savings shrink and can go negative. The converter's receipt shows the exact numbers for your payload.
Are the token counts exact?
Counts use the real o200k_base and cl100k_base tokenizers (the same BPE encodings OpenAI models use), so they are exact for those model families. Other vendors' tokenizers differ slightly — treat cl100k as an approximation for them.
Is the conversion lossless?
Yes for JSON and YAML: TOON encodes the full JSON data model, and decoding returns the original value. CSV is inherently untyped, so numbers and booleans are inferred when reading it; TOON to CSV only works for flat tabular data.
Can I convert TOON back?
Yes — TOON to JSON, YAML or CSV. The decoder is the official reference implementation, so it also validates your TOON and reports precise errors.
What's the file size limit?
10 MB. Everything happens in your browser's memory, so very large files depend on your machine — for bigger jobs use the TOON SDK directly in your pipeline.
Should my API return TOON instead of JSON?
No — TOON is a prompt-side optimization for LLM input. Keep your APIs on JSON and convert at the point where data enters a prompt.
Something else? Check What is TOON? or the cheat sheet.