llmweb-devproject

Building ToonifyIt - A Token-Efficient Data Format for LLMs

2026-07-106 min read

JSON is everywhere. It's the default data exchange format for APIs, configuration, and - increasingly - for feeding structured data into LLMs. But JSON is also verbose. Every key repeated, every bracket and quote adding tokens that don't carry meaning.

That's the problem ToonifyIt solves.

What is TOON?

TOON stands for Token-Oriented Object Notation. It's a compact representation of structured data designed to minimize token usage while preserving readability. Think of it as JSON stripped down to its essentials - the data without the ceremony.

The build

I built the conversion engine entirely client-side. No server, no API calls - everything happens in the browser. This was a deliberate choice: privacy matters when you're converting data that might contain API keys, user data, or proprietary information.

Key technical decisions:

  • Real-time conversion - as you type, the output updates. Debounced just enough to stay smooth.
  • Configurable delimiters - users can customize how TOON represents their data.
  • Token counting - the tool shows you exactly how many tokens you save. That's the whole point.
  • Copy/download workflows - one-click export for use in prompts or pipelines.

What I learned

Building a developer tool taught me that the UX bar is different from a marketing website. Developers don't want animations or surprises - they want speed, clarity, and keyboard shortcuts. Every interaction should be predictable.

The tool went live at toonifyit.com and has been getting steady traffic since. Not bad for a weekend project.

end of post