See inside your Parquet files.

Open a single file, select several at once, point at a whole folder, or paste an https:// URL to a remote Parquet file — it streams only the bytes it needs to read.

Walk the full schema — physical and logical types, nesting, repetition. Every column is tagged with its type and a fixed color, so the shape of the data reads instantly.

Min, max, distinct count, null ratio and compressed size for every column — read straight from the file’s metadata and row-group statistics. No full scan required.

Run custom SQL against any open dataset — joins, aggregates, CTEs and window functions. Results land in the same type-coded grid and export like any other slice.

Add predicate filters and multi-column sorts with type-aware operators. Row-group statistics skip chunks that can’t match, so it stays fast across files with millions of rows.

Take the filtered, sorted view and export just that slice — to CSV, JSON or a fresh Parquet file with the columns you kept. Everything is written locally.