CSV to JSON Converter

Convert CSV to JSON entirely in your browser, with nothing uploaded to a server. This page takes a spreadsheet, database, CRM, or analytics export and turns each row into a JSON object so you can seed an API, build a test fixture, import into MongoDB or another NoSQL store, or feed data straight to a frontend. The first row supplies the keys, every following row becomes an object, and the result is yours to copy or download.

What this conversion actually does (and does NOT do)

CSV is an untyped format: every cell is just text. JSON has real types (number, boolean, null), so the single biggest quality decision is type inference, meaning whether "42" becomes the number 42 and "true" becomes a boolean, or whether everything stays a quoted string. Both behaviors are legitimate. Numeric identifiers like leading-zero zip codes, phone numbers, and product SKUs should usually stay strings, because parsing them as numbers silently drops the leading zeros and corrupts the value.

There is no single "JSON" shape for a given CSV. Four outputs are all valid: an array of objects (the most common, one object per row), an array of arrays (raw 2D data), a column-keyed object (data grouped by column), and JSON Lines / NDJSON (one object per line, which is what MongoDB's mongoimport and many streaming pipelines expect). The first row normally supplies the keys; without a header row you can only produce arrays of arrays, not arrays of objects, because there are no names to map columns to.

The converter respects RFC 4180 edge cases that naive splitters break on: fields containing a comma or semicolon must be quoted, an embedded double-quote is escaped by doubling it (""), and a quoted field may legally contain line breaks, so splitting blindly on newlines would shred the data. The delimiter is not always a comma either. Many European exports and spreadsheets use semicolons (because the comma is their decimal separator) and TSV uses tabs, so the source must be parsed with the right separator or it collapses into one malformed column.

What it does NOT do is invent structure. Nested JSON is not native to flat CSV. Dot or slash header notation (user.name, address/city) and indexed columns are conventions a converter has to opt into, and arbitrary nested arrays of objects generally cannot be rebuilt from a flat table without explicit grouping logic. The conversion is lossless for structure but one-way-lossy for type intent: once a column is all strings, types have to be guessed, and converting JSON back to CSV will not necessarily restore the original quoting. Expect the JSON to be roughly 2 to 5 times larger than the source CSV, because every row repeats the key names and adds braces, quotes, and brackets, whereas CSV writes each header only once. NDJSON is a little smaller than a pretty-printed array and far more streamable for large files.

When to choose CSV to JSON vs keeping CSV

Choose CSV to JSON when the consumer wants structured data: an API request body, a JavaScript or test fixture, a document or NoSQL import, or a mock dataset. Keep the data as CSV, or go the other way with JSON to CSV, when the consumer is a spreadsheet, a BI tool, or a human reviewer scanning rows. If you mainly need to validate or pretty-print the JSON afterward, the JSON formatter is the next step.

How to convert CSV to JSON

1. Paste your CSV text into the input, or drop a CSV file onto the page. This page is locked to the CSV to JSON direction, so the opposite-direction button is hidden.
2. Set the "has headers" toggle. Leave it on so the first row maps to JSON keys, or turn it off to treat every row as data.
3. Click convert. The JSON appears in the output pane.
4. Copy the result to your clipboard, or download it as a .json file.

No upload. It runs on your device.

The conversion runs in Rust compiled to WebAssembly, directly in your browser. There are zero network requests during processing: open your browser's DevTools Network tab, or disconnect from the internet, and the tool still works. After the first page load it runs fully offline. Nothing is uploaded, so there are no temporary files on a third-party disk and nothing to delete afterward, because nothing ever left your machine. This matters because a CSV export often holds customer PII (names, emails, addresses), salaries, order history, or a proprietary database dump, and that is exactly the kind of data you should never paste into an unknown server. Here the bytes stay local and verifiably so.

No limits

No row cap, no file-size limit imposed by an upload, no watermark, no sign-up, and no ads. Convert one small CSV or several large ones back to back, all locally.

Frequently asked questions

How do I convert a CSV file to JSON without uploading it to a server?

Paste the CSV or drop the file on this page and convert. All parsing runs in WebAssembly inside your browser, so the file is read into local memory and never sent anywhere. You can confirm this in the DevTools Network tab, which shows no upload request during conversion.

Does this work offline without an internet connection?

Yes. Once the page has loaded, the Rust-to-WebAssembly engine is in your browser and needs no further network access. You can disconnect from the internet and the CSV to JSON conversion still works exactly the same.

Will my CSV with customer or sensitive data be uploaded anywhere?

No. Conversion happens entirely on your device with zero network requests, so a CSV containing emails, salaries, or order data is never transmitted to a server. Because nothing is uploaded, there are also no server-side temp files or logs to worry about.

How do I keep numbers as numbers (and booleans and nulls) instead of strings?

This depends on type inference, the decision to turn "42" into the number 42 and "true" into a boolean rather than keeping them quoted. Be deliberate about it: identifier-like columns such as leading-zero zip codes, phone numbers, and SKUs should stay strings, because parsing them as numbers drops leading zeros and corrupts the value.

How do I convert CSV to an array of objects vs JSON Lines (NDJSON) for MongoDB?

An array of objects wraps every row object in one JSON array, which is the most common shape for APIs and frontends. JSON Lines, or NDJSON, writes one object per line with no enclosing array, which is what MongoDB's mongoimport and many streaming pipelines expect; it is also slightly smaller and far more streamable for large datasets.

Does the first row have to be a header for CSV to JSON to work?

For an array of objects, yes: the header row supplies the property names that each column maps to. Without a header row there are no key names, so you can only produce an array of arrays (raw 2D data) rather than named objects. Use the "has headers" toggle to tell the converter which case applies.

How do I handle a semicolon-delimited or tab-delimited CSV?

The delimiter is not always a comma. Many European exports use semicolons because the comma is their decimal separator, and TSV files use tabs. The source must be parsed with the matching delimiter, otherwise the whole row collapses into a single malformed column instead of separate fields.

Can it handle commas, quotes, and line breaks inside CSV fields correctly?

Yes. The parser follows RFC 4180, so a field containing a comma or semicolon is read from its surrounding quotes, an embedded double-quote escaped by doubling ("") is unescaped to a single quote, and a quoted field may span multiple lines without breaking the row. Naive line-splitting corrupts exactly these cases, which is why the proper parser matters.

How do I create nested JSON objects from CSV columns?

Flat CSV has no native nesting, so nested output relies on conventions like dot or slash header notation (address.city or address/city) or indexed columns. Arbitrary nested arrays of objects generally cannot be reconstructed from a flat table without explicit grouping logic, so plan your headers if you need structure.

Why is my JSON file bigger than the original CSV?

JSON is typically 2 to 5 times larger than the source CSV because every row repeats the key names and adds braces, quotes, and brackets, while CSV writes each header only once. If size matters for a large dataset, NDJSON is a little smaller than a pretty-printed array and streams better.

Need the reverse direction? Use JSON to CSV, validate or pretty-print results with the JSON formatter, or browse all text and data tools. To see why this runs locally, read how to work with JSON privately in the browser.