Guide
Ask questions about a spreadsheet
Most tools that read spreadsheets hand the rows to a language model and hope the arithmetic comes out right. This one writes a query instead and runs it on your machine, against your actual data — so a total is a total.
Open a spreadsheet PDF, Word, PowerPoint, PNG, JPG, WebP, Text · XLSX, CSV, Parquet, JSONOpen the file
Excel, CSV, TSV, Parquet or JSON. It is loaded into a database that runs inside your browser — the rows are never uploaded. You see the columns it found and how many rows there are before you ask anything.
Ask in ordinary language
“Which month grew the most?”, “total revenue by region”, “how many orders were late”. A query is written for the question, run locally, and the answer is read back from the rows it returned.
Keep going
Each answer stays in the conversation, so the next question can build on it rather than starting over. Suggestions appear for what to ask next, drawn from what the data actually contains.
The rows never leave
Only the shape of the table goes out — column names, types, and a few sample values so the query can be written correctly. The data itself stays in your browser and the query runs there. Nothing is stored on our side: no table, no query log, no results.
I don't know SQL — can I still use it?
That is the point. Ask in plain language: which month sold most, what the total is per region. The SQL is written for you and run against your rows.
Can it analyse the data, not just read it?
That is the whole point of it. Totals, averages, outliers, the change between two columns, the rows that break the pattern — asked in a sentence instead of built as a formula, and computed rather than estimated.
How do I know the number is right?
The query that produced it is shown to you. You can see where the figure came from, and check it yourself if it matters.
Is the file uploaded anywhere?
The rows are not. Spreadsheets and CSVs are opened inside your browser, and what is sent is a short description of the table — its columns, their types, a few sample rows. The query that answers your question runs here, against the full file, which is why a million-row spreadsheet costs about two kilobytes of traffic.
Can it change my file?
No, and deliberately so. The table is read, never written, so nothing gets sorted, edited or saved over by accident. Answers come back as text and figures you act on yourself.
Which formats does it take, and which are free?
.xlsx, .xlsm, .xls, .csv, .tsv, .parquet and .json, straight from the file — no exporting, no converting, and no pasting rows into a chat box. All of them are free: a paid plan buys more files and more questions, not more file types.
How large a file can it handle?
Large enough for ordinary work. Very large files are read up to a row limit and the page tells you when that happened, so a truncated answer is never presented as a complete one.