AI Prep Tools

LLM JSON Repair & Schema Check

Repair malformed LLM JSON output, review every edit, and validate against a schema.

  1. Paste raw broken output from an LLM (including markdown fences or surrounding explanation).
  2. Review the 'What changed' list and toggle/undo individual repairs if needed.
  3. Optionally paste a JSON Schema to list path-level errors or copy a targeted fix prompt.
Loading local browser workspace...

Supported Formats, Limits & Accuracy

Syntax repair rules and supported draft 2020-12 schema checks are deterministic (Exact); truncation detection is Heuristic.

Supported:
Pasted text, .json, .txt
Processing:
Client-side browser execution ( how to verify)

Honest Limitations

  • When an LLM response was cut off mid-generation, repair only closes open strings/brackets — it never fabricates missing fields.
  • Schema check supports core JSON Schema draft 2020-12 keywords (type, properties, required, enum, items, minimum/maximum, pattern, oneOf/anyOf, additionalProperties); unsupported keywords are listed explicitly.

How to Use LLM JSON Repair & Schema Check

  1. Paste raw broken output from an LLM (including markdown fences or surrounding explanation).
  2. Review the 'What changed' list and toggle/undo individual repairs if needed.
  3. Optionally paste a JSON Schema to list path-level errors or copy a targeted fix prompt.

Frequently Asked Questions

Will this silently invent values to make a broken schema pass?
No. We strictly separate syntax repair from schema compliance. Missing required properties or wrong types are reported with JSON paths and a ready-to-paste fix prompt so you stay in control.

Next Steps & Related Tools