JSONL / NDJSON Viewer, Validator & Converter

updated 25 April 2026

100% client-side. Your files never leave this browser tab. No upload, no server, no tracking. Works offline once the page is loaded.

Before you start

You need one of the following:

• A .jsonl or .ndjson file (Newline Delimited JSON), or
• Raw JSONL text where each line is a valid JSON object.

This tool is designed to help you eyeball data without the lag of a full-blown IDE. It samples the first few thousand lines to keep the browser responsive. If your file is a giant 500 MB log export, don't worry — I use a streaming parser so we don't crash your tab, but the interactive table view is best suited for sets under 50,000 records.

Note that nested objects are flattened for the table view. If you have a key like "user": {"id": 1}, the column header will appear as user.id. This makes sorting and filtering much easier than digging through nested trees.

How to use it

1. Drop your .jsonl file into the zone above, or paste the text directly into the input pane.
2. Click Load to parse the records into the interactive table.
3. Type a query into the Filter box to find specific records (e.g., finding all "error" levels).
4. Switch the Mode between Substring (simple text match) or Expression (JS-like logic).
5. Use the Prev/Next buttons to navigate through pages of data.

Options explained

Substring vs. Expression Filter

Substring is the "safe" mode. It looks at every cell in a row; if the text you typed exists anywhere in that row, it stays. It's case-insensitive and great for quick searches.

Expression mode is for power users. It lets you write logic like score > 10 or status == "active". It treats the column headers as variables. If your headers have spaces or dots, wrap them in brackets or just stick to simple Substring matches.

The Pager

To keep the DOM from melting, I only render 50–100 rows at a time. The status bar tells you the total record count versus the filtered count. If you apply a filter, the pager resets to the first page of your results.

Example

Input (raw JSONL):

{"id": 1, "msg": "User login", "level": "info"}
{"id": 2, "msg": "Database timeout", "level": "error"}
{"id": 3, "msg": "User logout", "level": "info"}

Filtered View (Expression: level == "error"):

| id | msg              | level |
|----|------------------|-------|
| 2  | Database timeout | error |

Tips & common pitfalls

• Empty lines are ignored. Some exporters leave a trailing newline at the end of the file; I strip those automatically so you don't get "Invalid JSON" errors on the last line.
• Mixed schemas can be messy. If the first 10 lines have 3 columns but line 1,000 has 10 columns, the table will grow to accommodate the new headers. It's best if your JSONL is "rectangular" (consistent keys).
• Large values are truncated. If a cell contains a massive text blob (like a base64 image or a long stack trace), the viewer clips it to keep the row height sane. You can usually see the full text by hovering or using the Formatter tool.
• Case sensitivity only applies to Expression mode. Substring mode is always case-insensitive for easier searching.

Troubleshooting

Related tools

• JSONL ValidatorLine-by-line syntax check with jump-to-error.
• JSONL FormatterPretty-print or minify each record for better readability.
• JSONL to CSVFlatten nested objects and export to Excel → CSV.
• OpenAI ValidatorCheck chat-completion schemas before training.

See also: if you need to do something adjacent on this site, try Formatter to pretty-print or minify each JSONL record, JSONL to CSV to flatten JSONL into a CSV with dotted keys, or JSONL to JSON to convert JSONL into a JSON array (or back).

Frequently asked questions