llm-json is a lightweight Node.js tool that helps you get structured data from Large Language Models by:

A TypeScript SDK to extract and correct JSON from LLM outputs

Extract & validate JSON from messy LLM output. Designed for Claude / Codex CLI workflows where tool-use is unavailable. XML-tag aware, with multi-stage fallbacks and built-in repair.

Semantic compilation service for LokaScript - validates and compiles hyperscript from 24 languages, explicit syntax, or LLM JSON

A middleware router for processing and routing LLM JSON intents with Zod validation

Zero-dependency TS utility that repairs malformed LLM JSON output into valid, typed JSON.

Cross-platform streaming JSON parser for TypeScript optimized for LLM responses. Works on Node.js, Deno, Bun, browsers, and edge runtimes.

Streaming structured-render pipeline: incrementally parse incomplete LLM JSON into renderable cx component trees

Lightweight zero-dependency parser and validator for LLM JSON output

CLI Pulse — orienté agent/LLM (--json) : piloter le support (tickets) via l'API REST Pulse.

[llm-ui](https://llm-ui.com) JSON blocks for building custom components.

Token-Oriented Object Notation (TOON) – compact, human-readable serialization of JSON data for LLM prompts

Production-safe JSON repair and schema validation for LLM outputs

Parse partial JSON generated by LLM

micromark extension to support math (`$C_L$`, `\(C_L\)`)

Hardware accelerated language model chats on browsers

The official JSON schema converter for Valibot

Superfast runtime validators with only one line

Encode JSON → TOON (40-60% token savings for LLMs!)

Superfast runtime validators with only one line