JSONL Log Workbench

How to use

1. Load the JSONL

   Drop a file or paste the text. Invalid lines are listed separately with only their line number and reason.

2. Filter and redact

   Narrow the lines with a search term, name the keys and JSON Pointers to redact, then analyze.

3. Check and save

   Review the replaced values in the table, then save as JSONL or as a CSV of the columns you chose.

Scope of the result

Input is limited to 16 Mi characters and 100,000 lines, and each line to 96 levels of nesting and 250,000 nodes. The CSV takes 1–256 columns and the replacement text 1,024 characters. The table shows the first 200 valid lines. Keys you did not name, and anything embedded inside strings such as message, survive — this is not full anonymization. The CSV formula guard only prefixes a quote and cannot guarantee safety in every spreadsheet. Multi-line JSON, regular-expression search and streaming are not supported.

Frequently asked questions

Can it still work when some lines are invalid?

Invalid lines are listed apart from the valid ones. Saving stays disabled until you choose to drop them, so you don't ship a file with lines missing unnoticed.

Can I redact only the key named token?

Yes. Key rules match at every level and ignore case. Use a JSON Pointer when you want one specific location instead.

Is the redacted output safe to share as is?

Keys you did not name, and text written inside string values, remain. Review the output and add the entries you need.

Can objects and arrays go into the CSV?

If the value at the pointer is an object or array, it goes into one cell as JSON text. Whole tables are not flattened automatically.