TOON, or Token-Oriented Object Notation, is a compact, human-readable encoding for the JSON data model designed to reduce token usage when structured data is passed to language models. It combines indentation-based nesting similar to YAML with CSV-like tabular forms for repeated records, making regular collections especially compact while preserving a structure that models can follow.

The format is aimed at situations where applications need to place substantial structured context inside prompts without paying the full token cost of conventional JSON. It is particularly effective for arrays or keyed collections whose objects share the same fields, while still supporting nested objects and other common JSON-style structures.

The repository contains the TOON implementation and links to the broader specification and package ecosystem. It is useful for developers building LLM applications, agents, data pipelines, or prompt infrastructure where structured-data readability and context-window efficiency both matter.