---
title: How it works
description: Where the SQL you write and its results are handled, and how to connect to datasets
---

Whether you use Queria's web app, the CLI or Python, you can write the same SQL against the same data.

## Your SQL and its results

SQL you run in Queria's web app, the CLI or Python is computed on your computer. The results are not sent to Queria, and neither is the SQL itself. Only the CLI asks, the first time you use it, whether it may send your SQL, and sends it if you say yes (see [Telemetry](/en/telemetry#sql-you-run)).

Even for large tables, only the columns and rows your SQL uses are read. The whole table is never downloaded.

When you use Queria from an AI chat, Queria's server receives and runs the SQL the AI writes. What is recorded then is described in [Telemetry](/en/telemetry#mcp-server).

## Connecting takes the catalog URL

Each dataset is published in DuckLake format under `https://data.queria.io/<dataset>/`. When you connect from another tool, the catalog URL is all you specify.

The browser, the CLI and BI tools all read the same catalog, so the same SQL gives the same results everywhere, and you can hand your SQL to others as it is.

## How data gets published

Source data (open data distributed as CSV and other formats) has its encodings, column names and types standardized, and is then registered in the catalog with descriptions of its tables. How this is done is described in [Data quality](/en/data/quality).
