tooloora

CSV to JSON and back — in your browser

Convert CSV to JSON with delimiter detection, header recognition and optional typing — or flatten JSON back into CSV. Nothing is uploaded.

Runs locally — nothing is uploaded

Everything runs in your browser – no upload, no storage.

Convert CSV to JSON — and back again

Paste a CSV table and get JSON straight away: the delimiter is detected, the header row too, and numbers and booleans are typed on request. The reverse direction flattens nested objects into dot notation. A table preview and a per-row error list show immediately where a file goes wrong. Everything runs in your browser, with no upload and no sign-up.

What actually counts as a valid CSV file?

The reference text is RFC 4180 from October 2005, titled "Common Format and MIME Type for Comma-Separated Values (CSV) Files". It describes a surprisingly small core: fields separated by commas, lines terminated by CRLF, fields containing a comma, quote or line break wrapped in double quotes, and a quote inside a field doubled.

The document itself notes that no formal specification exists that everyone follows — and that shows in practice. German Excel versions write semicolons because the comma is taken as the decimal separator. Export tools use tabs. Some systems quote every field, others quote none. That is why this tool guesses the delimiter instead of prescribing one, and always lets you override the guess.

How is the delimiter detected?

Across the first 20 non-empty lines the tool counts how often comma, semicolon, tab and pipe occur outside of quoted fields. The winner is the character whose count is identical in as many lines as possible. Consistency beats frequency: a semicolon file with commas inside address fields is recognised correctly, because the comma count varies from line to line while the semicolon count does not.

Why do ZIP codes stay text?

Because 01067 would otherwise turn into 1067. Typing deliberately converts only what matches the JSON number format exactly:

  • 42, -3.5, 1e3 become numbers
  • 01067, +5, 1.234,00, 0049 30 123456 stay text
  • true and false (any capitalisation) become booleans
  • empty fields and null become null

That is the conservative reading: better one value converted too few than a product code, IBAN or phone number destroyed. If you want everything as text, simply switch typing off.

How do nested objects become columns?

CSV only knows flat tables, JSON does not. Going from JSON to CSV therefore uses dot notation:

JSONCSV column
{"address": {"city": "Berlin"}}address.city
{"tags": ["new", "premium"]}tags.0, tags.1 (or one column joined by ; )
{"note": null}empty cell

Column order follows first appearance, missing keys become empty cells. A dot inside a key name itself produces an ambiguous column name — the well-known price of this notation, and not something that can be resolved cleanly.

Honest limits

Everything runs in a single browser tab. Beyond a few megabytes typing gets sluggish, and above roughly 20 million characters the tool refuses to process at all so the tab stays usable. The delimiter guess works line by line and can be wrong when fields contain many line breaks — then set the delimiter manually. For gigabyte-sized exports, csvkit, jq or a small script are the right tools, not a browser tab.

Frequently asked questions

Is my file uploaded to a server?

No. Parsing and conversion happen entirely in your browser, the page works offline once loaded. That matters here: CSV exports are typically customer lists, order data or payroll extracts, and many online converters send every byte you paste to their backend. This tool has no upload endpoint at all.

How does the tool detect the delimiter?

It counts commas, semicolons, tabs and pipes outside of quoted fields in the first 20 non-empty lines and picks the character that appears the same number of times in as many lines as possible. Consistency beats frequency, so a semicolon file with commas inside free-text fields is still recognised correctly. You can always override the choice.

Why does my ZIP code 01067 stay a string when typing is on?

Because turning it into a number would destroy the leading zero. Typing only converts values that match the strict JSON number format, so 42 and -3.5 become numbers, while 01067, +5 and 1.234,00 stay text. true and false become booleans, empty fields become null.

What happens to nested objects when converting JSON to CSV?

They are flattened with dot notation: {"address":{"city":"Berlin"}} becomes a column named address.city. Arrays either get one column per item (tags.0, tags.1) or are merged into a single column, your choice. Empty objects, empty arrays and null become empty cells.

Is there a size limit?

Yes, an honest one. Everything runs in one browser tab, so files beyond a few megabytes make the page sluggish and above roughly 20 million characters the tool refuses to process at all. For gigabyte-sized exports a command-line tool such as csvkit or jq is the better fit.

Related tools

JSON Formatter & Validator — local, with error positionFormat, minify and validate JSON with exact error position, 100% in your browser.CSV to Excel and Excel to CSV — in your browserCSV to .xlsx and back, with fitted columns, bold header and protection against formula injection.Sort Lines & Remove Duplicates — local list cleanupSort lines, remove duplicates and empty lines, trim, reverse — with live statistics, no upload.