SQL Playground — run SQLite online, in your browser

SQL playground online, explained

SQL Playground is an online SQL editor that runs a real database engine inside your browser tab: SQLite 3.45.2, compiled to WebAssembly by sql.js 1.10.3. There is no database server behind it. Your queries, your .sqlite files and your spreadsheets are processed on your own machine and are never uploaded. Use it as a SQL sandbox for learning, to test a query before it goes into your app, or to explore a database file or CSV export someone sent you without installing anything.

It opens with the Chinook music-store sample already loaded, so your first query can run within seconds of the page loading.

What runs your queries

The engine is the official SQLite C library compiled to WebAssembly by the open-source sql.js project. The page loads two files from cdnjs.cloudflare.com: sql-wasm.js (about 50 KB) and sql-wasm.wasm (about 650 KB). After that every statement runs locally. Run SELECT sqlite_version(); to check the version yourself; the first query tab does exactly that.

The database lives in memory. Opening a file reads the whole thing into RAM, and changes stay there until you download the database or the playground saves it in your browser (see below). This build includes JSON functions (with -> and ->>), window functions, CTEs, math functions such as sqrt(), STRICT tables, RETURNING, FULL OUTER JOIN and FTS3/FTS4 full-text search. It does not include FTS5, the R*Tree module, generate_series or a REGEXP function, so those statements fail with "no such module" or "no such function".

What you can do in the playground

• Write SQL comfortably — syntax highlighting, autocomplete for your own tables and columns, bracket pairing, auto-indent, several query tabs and a one-click SQL formatter.
• See every result — run a whole script and each SELECT gets its own result tab; sort columns, copy the grid into Excel or Google Sheets, or export CSV, JSON, Markdown, Excel-ready CSV or INSERT statements.
• Chart a result — switch any result to a bar, line or area chart and download it as SVG.
• Understand a schema — the schema browser lists tables, views, indexes and row counts, and ◇ diagram draws an ER diagram from the foreign keys.
• Tune a query — 🧭 plan shows SQLite's EXPLAIN QUERY PLAN as a tree and flags full table scans.
• Fix errors faster — typos in table or column names get “did you mean …?” suggestions, and common MySQL, PostgreSQL and SQL Server syntax gets the SQLite equivalent.
• Edit without SQL — a spreadsheet-style grid for table data, plus buttons to create, rename and drop tables and columns.
• Keep your work — databases you change are kept in this browser (IndexedDB) and listed under recent; turn it off with one checkbox, or download a real .sqlite file.
• Example queries per sample — picking a sample database loads a query written for it, and 📚 examples lists 29 ready-made queries: eight for Chinook and seven each for Northwind, Employees and Shop.
• Share a query — 🔗 share copies a link that opens the same sample database and runs your query. Links never contain your own data.

The sample databases

The four samples are created by SQL inside the page, so they load instantly and work offline once the engine is loaded. They are small on purpose: you can see every row and check an answer by eye.

Chinook and Northwind are trimmed-down versions of the well-known teaching databases, not the full downloads: Chinook here has 19 tracks rather than several thousand. Shop is generated in your browser from a fixed random seed, so everyone gets exactly the same 7,705 rows and the exercises always have the same answers.

Query CSV, JSON and Excel files with SQL

You don't need a database file to start. Drop a .csv, .tsv, .txt, .json, .ndjson/.jsonl or Excel file (.xlsx, .xlsm, .xls, .ods) onto the page, or paste a URL and press fetch. The importer detects the delimiter and column types, shows a preview, and creates a real table; for workbooks you pick the sheet. Import several files and JOIN them in one query, then export the answer. Values with leading zeros, such as postcodes and product codes, stay text so nothing is silently changed.

1. Drop the file on the page (or use ⇪ import).
2. Check the table name and the detected column types, then press ✓ import table.
3. Query it: SELECT * FROM my_file WHERE amount > 100 ORDER BY amount DESC;

A column becomes INTEGER if every value is a whole number, REAL if all are numeric and TEXT otherwise. Postcodes, phone numbers and values over 15 digits stay TEXT, empty cells become NULL, and Excel dates arrive as YYYY-MM-DD text that sorts correctly. Untick detect column types to keep everything as text, or fix a column later with CAST. Excel files need the SheetJS reader, which is only downloaded the first time you open a spreadsheet.

Export formats

A result can be saved as CSV, JSON (an array of objects), a Markdown table, an Excel-ready CSV (UTF-8 with a byte-order mark and semicolons, so Excel opens it with accents and columns intact) or a file of INSERT statements for a table name you choose. You can also copy the grid and paste it into a spreadsheet. A chart downloads as SVG, and ⬇ db saves the whole database as a standard .sqlite file that opens in DB Browser for SQLite, the sqlite3 command line or any SQLite library.

Example SQL queries to try

Each example opens the right sample database and runs the query, so you can change it and see what happens.

SQLite cheat sheet

The statements you reach for most often, in SQLite syntax. Paste any of them into an empty database to try them.

SQLite vs MySQL vs PostgreSQL: differences that trip learners up

Most everyday SQL works everywhere, but a few things differ. If you are practising for MySQL or PostgreSQL, these are the spots to watch.

The biggest conceptual difference is typing. SQLite uses type affinity: a column declared INTEGER will still accept the text 'abc' unless the table is created STRICT. MySQL and PostgreSQL reject or convert such values. SQLite also has no separate DATE type: dates are stored as ISO text such as 2026-10-13 and handled with date(), datetime(), strftime() and julianday(). And foreign keys are only enforced after PRAGMA foreign_keys = ON;.

Which SQL it speaks

The engine supports joins of every kind including RIGHT and FULL OUTER JOIN, CTEs and WITH RECURSIVE, window functions, RETURNING, upserts, JSON and math functions, PRAGMA and ATTACH. Paging works with LIMIT 10 OFFSET 20 and the MySQL-style LIMIT 20, 10. PRIMARY KEY, NOT NULL, UNIQUE, CHECK and DEFAULT are enforced as usual. PostgreSQL-only syntax such as ILIKE, :: casts and JSONB is not available; the error message suggests the SQLite equivalent. The fundamentals you practise here (SELECT, joins, grouping, subqueries, CTEs, window functions) carry straight over to MySQL and PostgreSQL.

The editor runs as a script: statements execute in order and each one that returns rows gets its own result tab, while ▶ selection runs only the highlighted part. The layout works on phones and tablets too, though longer queries are nicer with a keyboard.

Saving your work, file sizes and limits

With keep my databases in this browser ticked (the default), a database you have changed is saved a few seconds later to your browser's IndexedDB storage (a database called jb-sql-playground) together with its query tabs, and listed under recent. The six most recent are kept; untouched samples and empty databases are not saved, and databases over 64 MB are not autosaved (download them with ⬇ db). Untick the box and nothing is kept after you close the tab. Your last 50 queries (under Query History), your practice progress and the recent list are kept in localStorage. Clearing your browser's site data removes all of it.

Because SQLite loads the whole database into memory, a few hundred MB is fine on a modern laptop, while multi-gigabyte files are better served by a desktop app such as DB Browser for SQLite. A file fetched from a URL is downloaded by your browser, so the host must allow cross-origin requests (CORS). Raw GitHub files and most open-data portals do.

There is no undo for data: DELETE, DROP TABLE and grid edits apply immediately. Wrap experiments in BEGIN; … ROLLBACK; or download a copy with ⬇ db first. The download keeps the connection open, so settings such as PRAGMA foreign_keys stay as they were. Grid edits run parameterised UPDATE, INSERT and DELETE statements, so a value like O'Brien is stored as text and never executed. Tables created WITHOUT ROWID can be changed with SQL but not in the grid.

Privacy: what stays on your device and what does not

Your tables, files and query results never leave the browser: the page has no server to send them to. It does make a few requests of its own, and you should know what they are:

• cdnjs.cloudflare.com: the sql.js engine on every visit, and the SheetJS spreadsheet reader (xlsx 0.18.5) the first time you open an Excel or ODS file. These are plain downloads; none of your data is attached.
• The URL you fetch: fetch and ?url= links download the file straight from that host, which sees an ordinary request from your browser.
• Google Fonts for the JetBrains Mono font, and Simple Analytics for a cookie-free page-view count.

A 🔗 share link contains the text of your query in the URL (and the sample name, if you used an unchanged sample) but never your data, so check the SQL itself for anything private before you send it.

Practise SQL with 32 exercises

The practice panel has 32 exercises on the sample databases: 10 beginner, 15 intermediate and 7 advanced, from a first SELECT to window functions and recursive CTEs. Your query is checked by comparing its result with a reference solution on a fresh copy of the data, so any correct approach passes; where the task asks for a sort order, the order is checked too. Hints and solutions are one click away, and your progress stays in this browser. A link such as ?practice=12 opens one exercise directly.

Keyboard shortcuts

More free tools