CSV to JSONL Converter

Convert CSV to JSONL (one JSON object per row) or JSON Lines / NDJSON back to CSV, with header, delimiter and type-inference options. Runs 100% in your browser — your data never leaves the page.

How CSV ⇄ JSONL conversion works

• CSV → JSONL: each CSV row becomes one JSON object on its own line, keyed by the header row (or col1, col2… without a header). Duplicate or empty header names are made unique.
• Type inference turns 42, 3.14, true/false and null into JSON types. Values like 007 or IDs longer than 2⁵³ stay strings so nothing is silently changed; empty cells stay empty strings.
• JSONL → CSV: the header is the union of top-level keys across all lines in first-seen order. Nested objects and arrays are written as compact JSON strings; null and missing keys become empty cells.
• Quoted fields, embedded delimiters, quotes and newlines are handled per RFC 4180. Use the CSV to JSON converter when you need a single JSON array instead, or JSON ⇄ JSONL to switch between array and lines.

What is JSONL (JSON Lines / NDJSON)?

JSON Lines — also called NDJSON (newline-delimited JSON) — is a text format where every line is one complete, self-contained JSON value, usually an object. There are no commas between records and no surrounding [ ]. Files typically use the .jsonl or .ndjson extension.

{"id": 1, "event": "signup", "plan": "free"}
{"id": 2, "event": "upgrade", "plan": "pro"}

JSONL vs a JSON array

A JSON array must be parsed as a whole, so a single missing bracket breaks the entire file. JSONL can be streamed, appended to, split and processed line by line (head, grep, wc -l), and a corrupt line only affects that one record. Use a JSON array for small API payloads and config; use JSONL for large datasets, logs and anything written incrementally.

Where JSONL is used

• OpenAI fine-tuning datasets and Batch API input/output files (one request or example per line).
• BigQuery, Snowflake, Redshift and Athena bulk loads (NEWLINE_DELIMITED_JSON).
• Structured logs from Docker, Kubernetes, Bunyan, Pino and most log shippers.
• Elasticsearch / OpenSearch bulk API, MongoDB mongoexport, Hugging Face datasets.
• Data pipelines with jq -c, Spark, pandas read_json(lines=True) and DuckDB.

All JSONL tools on DevToolsDaily run 100% in your browser — your data is never uploaded. See also the JSONL formatter, JSON ⇄ JSONL converter and CSV ⇄ JSONL converter.