CSV SQL Query Tool

Run SQL on a CSV Without Importing It Anywhere

The usual paths from "I have a CSV" to "I have an answer" all have a toll: import it into a database you have to stand up, upload it to someone's server, or wrestle it in a spreadsheet. This tool takes a fourth path: a real analytical database (DuckDB, the engine behind a wave of modern data tooling) compiled to WebAssembly and running inside your browser tab. Your file is read locally with automatic type detection, becomes a table, and answers real SQL: GROUP BY, CTEs, window functions, string and date functions. Close the tab and everything is gone. Nothing is uploaded, which matters when the file is a customer list or a payroll extract. Before you query unfamiliar data, the Data Profiler gives a column-by-column shape report worth a first look.

Joins: the Queries Spreadsheets Can't Do

The feature that earns this tool a bookmark is the second file. Drop in orders.csv and customers.xlsx and you have two tables; a one-line JOIN replaces the VLOOKUP gymnastics that consume entire afternoons, and the tool even suggests the join when two tables share a column name. Excel workbooks import each sheet as its own table, and Parquet files (increasingly what data teams hand around) load natively; the Parquet File Explorer remains the deeper viewer for Parquet metadata and schemas. When your prototype query needs to become a real table somewhere, the CSV to SQL Generator turns the same file into CREATE TABLE and INSERT statements, and the CSV & Delimited Text Converter handles format conversion without the database.

Plain-English Questions, Visible SQL

If SQL is not your daily language, type the question instead: "total revenue by region for shipped orders." What gets sent to generate the SQL is the question plus your column names and types, never the data itself (you can opt in to including three sample values per column when a question needs them). The generated SQL appears in the editor, not in a black box: you read it, adjust it, and run it yourself, and a deterministic check flags any table or column the query mentions that does not actually exist in your files. That visibility is deliberate. The fastest way to learn the SQL for your own data is to see the query that answers your own question.

From One Query to a Reporting Pipeline

An ad-hoc query is the prototype of a report somebody will want every week. I build the production version: scheduled extracts from ERP, POS, and operational systems, joins across sources that do not naturally talk, and reporting your team opens instead of rebuilding the same spreadsheet. See reporting and analytics services, ETL and data pipelines, or try the interactive BI dashboard demo.