CSV to JSON Converter

What the parser does with your CSV

The converter first splits your text into lines and drops completely empty ones. It then decides on a delimiter. With auto-detect on, it counts commas, tabs, semicolons and pipes in the first line only and uses whichever occurs most. A tie goes to the earlier candidate in that order, and a one-column file with no delimiters defaults to comma. Each line is then read character by character, tracking whether it is inside double quotes. Inside quotes, a doubled "" becomes one literal quote, and delimiters are ignored.

Output shapes

• Array of objects: the header row supplies the keys. If you untick "First row is a header", keys are generated as column_1, column_2 and so on.
• Array of arrays: each row is a plain list. The header row is dropped when the box is ticked, otherwise it stays as a row.

Worked example

A semicolon file, as exported by Excel in locales that use a decimal comma:

name;city;age
Ada;"London, UK";36
Grace;NYC

Auto-detect counts two semicolons against none of the others, so it picks ;. Object output is:

[{"name":"Ada","city":"London, UK","age":"36"},{"name":"Grace","city":"NYC","age":""}]

The quoted comma survived, 36 is the string "36" because nothing is type-converted, and Grace's missing third cell became an empty string. In array-of-arrays mode the same data gives [["Ada","London, UK","36"],["Grace","NYC"]], and the short row stays short.

Edge cases worth knowing

• Numbers stay strings. Convert them in your own code, or the receiving API may reject "36" where it expects 36.
• Duplicate header names collapse. With two columns named a, the later one overwrites the earlier in object output.
• Extra cells are dropped. A row longer than the header loses its overflow in object mode. Array mode keeps everything.
• Quoted line breaks are not supported. The text is cut into lines before quotes are considered, so a multi-line cell will split into separate rows. Delimiters inside quotes also count toward auto-detection on the first line, so set the delimiter by hand if your header row contains quoted text.
• Pretty-print indents with two spaces. Untick it for single-line output, which is smaller to paste into a request body.

To go the other way, use JSON to CSV. To pull out specific columns before converting, try the CSV column extractor, or turn the data into a document table with CSV to Markdown.