JSON to SQL Converter

No. Your JSON is never sent to our servers - all parsing, type detection, escaping, SQL assembly, copy, and download happen directly in your browser on your device. The only network request the tool ever makes is the one you trigger explicitly with URL Import, which goes from your browser to the URL you enter.

The input must be a non-empty JSON array of objects, like [{"id":1,"name":"Alice"},{"id":2,"name":"Bob"}]. The first object's keys are NOT the only ones used - the converter walks every object and collects the union of all keys, so objects with missing fields still produce columns with NULL values.

For each column, the worker scans every non-null value across all objects. If they are all integers it picks INTEGER. If they are all numbers (including decimals) it picks REAL. If they are all booleans it picks BOOLEAN. Otherwise it picks TEXT. Empty strings and null/undefined are ignored during type detection so a column with a few missing values still gets the correct type.

String values are wrapped in single quotes and any embedded single quotes are escaped by doubling them (''). Numbers stay unquoted, booleans become TRUE/FALSE keywords, and null/undefined values become the SQL NULL keyword. The output follows standard SQL escaping rules so it pastes cleanly into PostgreSQL, MySQL, SQLite, SQL Server, and any other SQL tool.

The Upload picker accepts .json, .jsonl, and .txt files. Drag-and-drop accepts the same. URL Import fetches the file directly from your browser to the target URL - the target site must allow cross-origin requests. You can also type or paste JSON directly into the textarea.

When Batch INSERT is on, the worker generates a single multi-row INSERT statement with the (...), (...), (...) tuple syntax instead of one INSERT per row. Batch mode is significantly faster for large bulk loads in PostgreSQL, MySQL, and SQLite because it reduces transaction overhead and avoids parser re-initialization for each row.

Your settings (table name, schema prefix, batch mode, empty value handling) are saved to your browser's local storage under the key "json-to-sql-converter" and restored on the next visit. Your JSON input and generated SQL output are not saved - storing large files on every keystroke would slow the tool down. Re-paste, re-upload, or re-fetch your JSON after a refresh.

The tool is optimized for big files. The full SQL output is always generated and available via Copy and Download, but the on-screen displays are capped: the input textarea is limited to the first 500 lines and the SQL output preview is also limited to the first 500 lines. The actual parsing and SQL generation runs in the background so the page stays responsive throughout.

Yes. Set the Schema field to your schema name (e.g. dbo, public, mydb). The generated SQL will produce fully qualified references like `public`.`my_table`. Leave it empty to omit the schema prefix entirely. Identifier quoting uses MySQL-style backticks.

Parsing the JSON, scanning columns for type detection, escaping every value, and assembling thousands of INSERT statements all happen in the background without blocking the page. The spinner overlay animates the whole time, you can still scroll and click other tabs, and the browser never marks the tab as unresponsive. This is what makes the tool work on multi-megabyte JSON files.

For each column, the converter scans every non-null value across all objects in the array. If they are all integers it picks INTEGER. If they are all numbers (including decimals) it picks REAL. If they are all booleans it picks BOOLEAN. Otherwise it picks TEXT. Empty strings, null, and undefined are excluded from type inference so a column with a few missing values still gets the correct type.

Different objects in the JSON array can have different keys. The converter walks every object and collects the union of all keys, so missing fields still produce columns with NULL values in the affected rows. You don't need to normalize your JSON before converting.

A complete CREATE TABLE statement is generated with one column per detected key, identifier quoting with backticks, and the auto-detected SQL type. The table name comes from the Table field (default my_table) and is also backtick-wrapped to handle reserved words and special characters safely.

Set the Schema field to a database, namespace, or schema name (e.g. dbo, public, mydb) and the generated SQL will produce fully qualified references like `public`.`my_table`. Leave it empty to omit the schema prefix entirely. Trailing dots are stripped automatically.

Toggle Batch INSERT in the output toolbar to combine all rows into a single multi-row INSERT statement using the (...), (...), (...) tuple syntax instead of one INSERT per row. Batch mode is significantly faster for large bulk loads in PostgreSQL, MySQL, and SQLite because it reduces transaction overhead.

Use the Empty values dropdown to choose how blank or whitespace-only string values are written: NULL inserts the SQL NULL keyword (default), Empty string '' inserts a single-quoted empty string. Pick based on whether your column is nullable or NOT NULL DEFAULT ''.

Numbers are emitted unquoted, booleans become TRUE/FALSE keywords, null/undefined become the SQL NULL keyword, and string values are wrapped in single quotes with embedded apostrophes doubled (' -> ''). Standard SQL escaping that pastes cleanly into PostgreSQL, MySQL, SQLite, SQL Server, and any other SQL tool.

Four ways to load JSON: type or paste straight into the textarea, click Upload to pick a .json / .jsonl / .txt file, drag a file onto the Input area, or click URL Import to reveal a URL field that fetches the file directly from your browser. URL Import requires the target site to allow cross-origin requests.

When you load a large JSON file, only the first 500 lines are shown in the input textarea. The full file is still loaded and used during conversion - the cap is purely a display safeguard to keep the page responsive when the source is multi-megabyte. A notice next to the Input label tells you when the preview is limited and shows the full file's row and column count.

The output panel shows the generated SQL with color-coded syntax highlighting: keywords in blue, identifiers in amber, single-quoted strings in green, and numbers in purple. The displayed code is limited to the first 500 lines for responsiveness, and a notice next to the Output label tells you the preview is limited. Copy and Download always work on the full untruncated output.

When you load a file or fetch a URL, a spinner overlay appears over the Input area with a status message. When you click Generate SQL, a spinner overlay appears over the Output area until the conversion finishes. This makes it obvious when the tool is working rather than appearing frozen.

Click Copy SQL to copy the entire generated output to your clipboard. The button label briefly changes to "Copied!" for 1.5 seconds as confirmation, then reverts.

Click Download SQL to save the output as a .sql file named after your table (for example my_table.sql). The blob is created locally on your device and the temporary object URL is released after the download is triggered.

Your settings (table name, schema prefix, batch mode, empty value handling) are saved in your browser and restored on the next visit. Your JSON input and generated SQL output are not saved - they can be very large, and persisting them on every keystroke would slow the tool down. Re-paste, re-upload, or re-fetch your JSON after a refresh.

Click Clear to wipe the input textarea, the generated SQL, the URL Import bar, and any error message in a single action. Settings (table name, schema prefix, batch mode, empty value handling) are reset to defaults and the persisted localStorage entry is removed.