JSON to CSV converter

1. Choose JSON or JSON Lines / NDJSON. Paste your data or open a local .json, .jsonl or .ndjson file up to 2 MB in the matching mode.
2. Preview the input. For a wrapped JSON response, choose its records array from the discovered paths. Keep nested fields in a cell or flatten objects, then select a delimiter.
3. Check the rows and columns and download the CSV. The export includes every accepted record, not just the preview rows.

For {"data":[{"id":"001","name":"Ada"},{"id":"002","name":"Lin"}],"total":2}, preview once, then choose /data under Records to export. The CSV has two records with id and name columns; total is response metadata and is excluded. For a log export, choose JSON Lines / NDJSON and place each object on its own line.

JSON mode accepts one object or an array of objects. It discovers up to 100 arrays containing only objects below response object properties, including paths such as /response/items. Select an array explicitly; the tool does not guess which response data you need. Arrays inside other arrays are not expanded.

JSON Lines / NDJSON mode reads one complete object per line and ignores blank lines. Errors show the original line number. Pretty-printed multi-line objects, JSON comments, null, primitive values and arrays on a line are not accepted. CSV input is not supported.

Preserve mode stores nested objects and arrays as JSON inside a cell. Flatten mode expands objects with JSON Pointer-style paths: / means a level, ~1 escapes / in a key, and ~0 escapes ~. Arrays remain in one cell; rows are never multiplied.

Missing fields and null become empty cells. CSV does not retain JSON types; spreadsheet software may infer dates, remove leading zeros or convert long identifiers. Import those columns as text.

Numbers outside JavaScript's safe range are rejected instead of silently rounded. Quote large identifiers in the original export. Decimal calculations follow JavaScript number precision.

Limits: 2 MB input, 10,000 exported objects, 200 columns, 200,000 cells, 64 overall nesting levels (per input object for JSON Lines) and 12 nested object levels in flatten mode. Spreadsheet text protection is on by default and prefixes risky text with an apostrophe; this changes those values intentionally.