CSV JSON Converter

Paste your CSV or upload the file, keep "first row is header" turned on, and the converter reads each remaining row as one object whose keys come from that header line. The result is a JSON array of objects, the same shape most APIs expect for a list of records. If you turn the header option off, you get an array of arrays instead, preserving raw row order without named keys. Blank header cells fall back to predictable names like column1 and column2 so nothing is dropped. The delimiter is auto-detected from the first line, or you can set comma, semicolon, tab, or pipe yourself, and you can pretty-print with 2, 4, or 8 spaces of indentation. Paste your data above and click convert to get clean, copy-ready JSON in one step.

CSV has no concept of data types, so every cell arrives as plain text. If type inference is switched off, the converter plays it safe and keeps each value exactly as written, which means 42 and true become the strings "42" and "true" in your JSON. Turn type inference on and the parser promotes whole numbers to integers, decimals like 3.14 to floats, true and false to real booleans, and empty, null, or n/a cells to JSON null. That gives you output your code can use directly without a second cleanup pass. The trade-off is that values such as zip codes or IDs with leading zeros may lose them once read as numbers, so leave inference off when you need every character preserved. Toggle the type inference option above to match how your downstream tool expects the data.

A flat CSV table cannot hold a nested structure directly, so the converter offers a flatten option that maps nested keys into dot notation. An object like {"address": {"city": "NYC"}} becomes a single column named address.city, and deeper nesting keeps chaining with more dots. Arrays inside a record are preserved as JSON strings packed into one cell rather than spread across columns, so no data is silently lost. The tool scans every object first, collects the complete set of keys in the order they first appear, and writes one column per key, so records missing a field just get an empty cell instead of breaking row alignment. With flatten turned off, nested objects are written as JSON text in their column. Pick the output delimiter you need, enable flatten if your JSON has nested objects, and convert to a tidy table.

When auto-detection is on, the converter inspects the first line of your CSV and counts how often commas, semicolons, tabs, and pipes appear, then picks whichever is most common as the separator. That works well for ordinary exports, but it can guess wrong on files where a quoted field happens to contain many commas or where the header is unusually short. Setting the delimiter manually removes that ambiguity: you tell the tool exactly which character splits your columns, including space-separated data that auto-detection might not expect. European exports often use semicolons because the comma is a decimal mark, while tab-separated files are common from spreadsheets, so manual control matters across regions. If your first conversion looks misaligned, switch from auto to the exact delimiter your file uses and convert again for correct columns.

The parser follows the RFC 4180 convention used by spreadsheets and most data exports. Any field wrapped in double quotes is treated as a single value even when it contains the delimiter or a line break, so a cell like "Smith, John" stays in one column instead of splitting into two. A doubled quote inside a quoted field, written as "", is read as one literal quote character rather than ending the field early. This is what lets addresses, descriptions, and multi-line notes survive a conversion intact. If a quote is opened but never closed, the tool reports it as an error so you can fix a malformed file rather than receiving silently wrong output, and rows with an unexpected column count are flagged as warnings. Paste a tricky CSV above and the column info tab will show you exactly how each field was parsed.