Divide JSON Lines by record count
Paste the stream, enter Records per part, and choose Split records. Each output file contains complete records; a JSON object or array is never split across parts. Download files individually or choose a file in Preview output part to inspect and copy that part.
Example: three records in two files
The example contains id 1, id 2 and id 3 on separate lines. The default two records per part creates part-001.jsonl containing the first two records and part-002.jsonl containing the third. Both downloads end with LF. Changing the count clears every old preview and download.
Choose a practical part size
Records per part must be an integer from 1 to 10,000. A maximum of 100 parts prevents the page from creating thousands of links and browser object URLs. Increase the part size when the requested split exceeds that limit. Parts are based on record counts, not byte sizes: one large object can make one part much larger than another.
Validation covers the whole input before any parts are offered, so a late malformed line cannot produce a misleading partial success. The current-output copy button copies only the selected part. There is no automatic multi-file download or ZIP archive.
Use JSON Lines Viewer to inspect records before splitting. JSON Lines Merger can append two selected parts later, preserving their chosen order.
Input and numeric limits
Processing is local and edits clear stale results. JSON Lines inputs accept LF and CRLF, with one optional final newline. Blank records, a leading BOM and standalone CR are rejected. Browser textareas may normalize pasted line endings before the parser receives them. Pretty-printed multiline objects must first be put on one line per record.
The reader rejects duplicate object keys, nonfinite numbers, unsafe integers, and nonzero numbers such as 1e-400 that become zero at JavaScript precision. Other decimals still use floating-point arithmetic; negative zero and numeric spelling can normalize. Store exact identifiers or precision-sensitive decimal text as strings. Each record is limited to 64 nesting levels; input is limited to 200,000 characters (combined when merging), 10,000 records and 20,000 values. Output is capped at two million characters. Input stays in memory and is lost on reload.





