JSON to CSV Converter

Convert a flat record array

1. Choose Pasted JSON or Local file. Files must use UTF-8, have a .json or .txt suffix and fit 2 MiB. The source must be a nonempty array of objects whose values are strings, numbers, booleans or null. Nested values and JSON Lines are not supported.
2. Choose comma, semicolon or tab. Prefix formula-like cells is off by default: enabling it changes matching headers and values, including negative numbers, and does not guarantee spreadsheet safety. Use example fills a two-record source without starting conversion.
3. Select Convert and read the data-row, column and output-byte counts. Conversion details shows how records become first-seen columns and quoted text. Copy and Download use the full current output even when the default preview is shortened.

Understand the table and its limits

Columns follow each decoded key's first appearance while reading records and their written members in order. Numeric-looking keys are not sorted. Later keys add columns and earlier rows receive empty cells there. Empty objects remain rows if another record supplies a column; a document with no columns cannot convert.

Strings retain their decoded characters. Number text such as 9007199254740993, 1.2300, 1e+03 and -0 keeps its original spelling when prefixing is off; it is not evaluated as a browser number. Booleans become true or false. Null, missing members and empty strings all produce empty CSV cells, with separate counts in the result.

CSV is not reversible JSON and does not preserve types. A number and the same text in a JSON string can produce identical cells. Spreadsheet software may change leading zeros, large numbers, dates or formulas even when every field is quoted. Choose matching separator and text-import settings rather than assuming double-click opening will preserve values.

Every output field is quoted and embedded quotes are doubled. Records end with CRLF, including the last; embedded CRLF, LF, CR and tab remain inside their fields. Output is UTF-8 without a BOM. Semicolon and tab are explicit dialects, not promises that every CSV or TSV reader handles their quoting.

Follow the supplied records

The supplied 86-byte example becomes two data rows and five columns in first-seen order: 10, 2, name, new and flag. With comma and prefixing off, the output is 97 UTF-8 bytes including a header and final CRLF. The large identifier and 1e+03 stay literal, and the comma and quotes in the name stay within one field.

The same example contains four missing members, one null and no empty-string values. Those five cells all look empty in CSV, but their causes remain visible in the counts. One field, -0, matches the formula-like heuristic. Prefixing changes it to '-0 and makes the output 98 bytes; that is a documented data change, not a safety guarantee.

An empty object followed by a record containing a: 1 produces header a, an empty first data cell and then 1. An empty array or only empty objects fails because no table columns exist. Two differently escaped spellings of the same decoded key in one record also fail instead of overwriting a value.

Explicit bounds and unsupported data

Original files and raw pasted UTF-8 are limited to 2097152 bytes; raw or decoded source also fits 2097152 UTF-16 units. The table allows 10000 data rows, 256 union columns and 250000 data cells counted as rows times columns, including empty cells. A key, decoded string or number token is limited to 65536 Unicode scalar characters before prefixing.

Output is limited to 4194304 UTF-8 bytes including quotes, prefixes and final CRLF. A worker taking more than 15000 milliseconds is terminated with no partial export. The ordinary preview shows at most 100000 Unicode scalars; Copy, its full readonly fallback and Download use the complete accepted output.

A key cannot be empty or only ASCII space, tab, CR or LF. Duplicate decoded keys, unpaired surrogates and decoded ASCII controls other than tab, CR and LF are rejected; DEL is also rejected. These are deliberate tabular interoperability restrictions, not a claim that every rejected source violates JSON grammar. One initial signature is accepted, but a second outside JSON text is not ignored.

Prefixing prepends one apostrophe to a cell beginning with tab, CR or LF, or whose first character after leading ASCII spaces, tabs, CR and LF is =, +, -, @ or their full-width counterparts. Original whitespace stays. The rule applies to headers too; if it creates duplicate headers, conversion fails. It is not universal sanitization and other software may change or undo its effect.

Local files and current output

This tool adds no document upload, account, saved history, source logging or data-bearing share URL. Editing, changing sources or options, replacing a file, cancelling or clearing retires the old output. Existing site services remain subject to the site's policies; local processing is not certification of those services or deployment.