I built Recast because I got tired of fixing the same broken JSON over and over. Whenever I (or tools like n8n) ask an LLM for structured data, I keep getting almost-correct output: single quotes, trailing commas, Python-style True/False/None, markdown fences, random prose, or truncated responses. Any of these make the JSON invalid, and most programs just reject it. Existing “JSON fixers” usually felt like black boxes — they either silently changed things or gave no explanation of what they did. For anything that ends up in real workflows, that lack of visibility was a dealbreaker. So I wanted something that: Actually explains every change in plain English Runs entirely in the browser (nothing is uploaded) Makes it clear when nothing was dropped The core idea stayed the same throughout: honesty + reliability + privacy. Everything else (batch mode, schema support, history, compare view, etc.) grew from using it myself on real messy outputs. Would love your feedback — especially if you work with AI-generated JSON or n8n workflows.

Recast Pro

Messy AI output → clean, valid, explained JSON

Messy AI output → clean, valid, explained JSON