JSONL Lens

a fast viewer for jsonl and ndjson files of any size. records are indexed by byte offset in one streaming pass and read on demand, so a multi-gigabyte training set is on screen while its index builds and the file is never loaded into memory. the viewer is free: grid, find with regular expressions, schema tab, record detail. pro, a one-time purchase, adds whole-file filter queries, export, exact stats and grid themes; try it for 14 days before you buy.

Free - the full viewer, for files of any size

• Any size: streaming line index with live progress; browse records while indexing runs. A 1.02 GB, 5,000,000-record file indexes in 3.7 s.
• Find in file: full-text search across every record, or a regular expression tested against each record on its own; case toggle, next/previous jump between matching records - built on the same streaming scan, so it works on multi-gigabyte files.
• Grid view: records as rows, fields as columns - column set derived from a sample of the file, nested fields as dot-paths like user.plan.
• Schema tab: every field path seen in a sample of the file, with present, missing and null counts, types, min, max, average, distinct strings and example values.
• Record detail: the selected record pretty-printed with key order and int64 precision preserved; copy is one click.
• Robust: malformed lines, blank lines, CRLF endings and oversized records are handled and reported, never fatal.
• Reads in place: the file is never copied or converted; the index is a list of byte offsets kept in memory and thrown away when the tab closes.
• Private by construction: runs locally. no account, no telemetry. the free viewer makes no network calls; the only network traffic is the license activation, and only after you paste a key.

Built for machine learning and data engineering workflows: training data, eval sets, model outputs, exports and logs.

Pro, optional one-time purchase

• Whole-file filter queries: level == "error" && status >= 500, streamed over every record; the grid becomes the matching set.
• Export: all records, filter matches or search matches, to jsonl or csv, with a cancellable progress bar. never over the source file.
• Exact stats: the schema tab's full scan reads every record instead of a sample.
• Themes: eight looks for the grid: default, hacker, cyberpunk, goblincore, neon, nosferatu, halloween and blue lagoon, each tuned for light and dark editors. run JSONL Lens: Pick theme to try one; trying is free for the window, keeping it is pro. the default look never changes.

Pro starts with a 14-day trial on each machine; no key needed to try it.

$49 personal, $99 business, one-time, includes 12 months of updates. 14 day refund. a personal key activates on up to 3 machines, a business key on 10; release a machine any time from the Polar customer portal.

buy on polar (personal, business), then run JSONL Lens: Enter license key. that makes one activation call. nothing else leaves the machine.

the free viewer is the complete viewer, not a demo. everything that was free in 0.1.0 stays free in every later build; pro adds the four items above and nothing is taken away.

Part of the Lens family

• Parquet Lens for VS Code: the same table, search and schema panes for Parquet, Arrow and Feather files.
• On JetBrains IDEs: JSONL Lens, Parquet Lens and Log Lens - one engine, a shared filter grammar, exports that open in each other.