Convert JSON to CSV

CSV is a flat, two-dimensional format: one header row of column names, then one row per record. JSON is a tree. The conversion only makes sense when your JSON is a list of records with the same rough shape — a JSON array of objects — where each object maps naturally onto a row and each key onto a column. JSON Copilot takes that array, walks every object once to collect the column names, then emits one CSV line per object.

Concretely, an input like [{"id":1,"name":"Ada"},{"id":2,"name":"Alan"}] becomes a header of id,name followed by 1,Ada and 2,Alan. The column order follows the order keys first appear while scanning the array, not alphabetical order, so the CSV reads the way your objects are written. Numbers and booleans are written as-is (1, true), and strings are written literally, only getting quotes when they actually need them.

If you pass a single object instead of an array, it is treated as a one-row table. If you pass an array of plain values such as ["a","b","c"], there are no keys to become columns, so the output is a single column named value with one row per element. That keeps the tool predictable no matter exactly how your data is shaped.

Real JSON is rarely perfectly rectangular. Two things make it messy: objects that do not all share the same keys, and values that are themselves objects or arrays. JSON Copilot handles the first with a union of keys — it scans every object in the array and records each distinct key the first time it sees it. The resulting header is the superset of all fields, so a key that appears in only one record still gets its own column, and records that lack it simply have an empty cell there.

The second problem — nested values — cannot be flattened without inventing a convention, so the converter keeps them intact as compact JSON. A value like ["admin","editor"] is written into its cell as the text ["admin","editor"], and {"name":"Crypto","size":4} becomes that same JSON string. Because those strings contain commas and quotes, the whole cell is wrapped in double quotes and the inner quotes are doubled, so a CSV parser on the other end still reads it as one field.

This is a deliberate trade-off: you get lossless output — nothing is dropped and no keys are silently merged — at the cost of cells that hold JSON rather than fully exploded columns. If you need each nested field in its own column, flatten the JSON first (the editor’s flatten tool turns nested keys into dotted paths) and then run the CSV conversion on the flat result.

The most common reason to convert JSON to CSV is to get data out of an API and into a spreadsheet. Excel, Google Sheets, and Numbers all import CSV natively, so a paginated API response or a saved JSON export can become a sortable, filterable table in seconds. Analysts who do not write code can then pivot, chart, and annotate the data without ever touching the original JSON.

CSV is also the lowest-common-denominator import format for business-intelligence and database tooling. Warehouse loaders, BI dashboards, and bulk-import screens almost always accept CSV even when they cannot parse arbitrary JSON. Converting first means you control exactly which columns exist and in what order, rather than relying on another system’s opinionated JSON flattening.

It is handy for one-off data work too: deduping a list, diffing two exports, sharing a subset of records with someone who lives in spreadsheets, or turning event JSON into rows you can quickly eyeball. Because the conversion is client-side, you can do all of this with data you would never want to paste into an online uploader.

A few inputs behave in ways worth knowing. An empty array has no rows to write, and an array of objects that all have no keys has no columns, so both are reported as errors rather than producing a blank file. If your array mixes objects with non-objects — say an object and a bare string — the converter falls back to the single value column mode instead of guessing a header.

Missing keys never shift your columns. Because every row is written against the full union of keys, a record that omits a field gets an empty cell in that position, keeping the grid aligned. Null values are written as empty cells too, while false and 0 are written literally as false and 0 — they are real values, not blanks.

Quoting follows standard CSV rules so the file survives a round trip. Any cell containing a comma, a double quote, or a newline is wrapped in double quotes, and literal double quotes inside are escaped by doubling them. That is why nested JSON cells look like "{""name"":""Crypto""}" in the raw file yet import as a single clean field. When in doubt, run the result back through the CSV to JSON tool to confirm it parses the way you expect.