JSON to SQL INSERT Converter

API responses, exports and fixtures often arrive as JSON, but the database wants SQL. Paste an array of objects and get INSERT statements — batched into multi-row statements — plus an optional CREATE TABLE whose column types are inferred from the values: integers, decimals with the right precision, booleans, dates, text and JSON. Objects that lack a key get NULL; nested objects and arrays are stored as JSON or flattened into dot-named columns. Identifiers, strings and booleans are written correctly for MySQL, PostgreSQL, SQLite, SQL Server or Oracle.

How to use this tool

1. Paste a JSON array of objects (one object or {"data": […]} also works).
2. Enter the table name and choose the database.
3. Choose whether nested objects become JSON columns or flattened columns.
4. Optionally include CREATE TABLE or DROP TABLE, or name a key for upserts.
5. Copy the SQL.

Formula

columns = every key in any object (first-seen order) · one VALUES tuple per object · missing key → NULL

Types follow the values: whole numbers → INTEGER (BIGINT beyond ±2,147,483,647), other numbers → DECIMAL(p,s) from their digits, true/false → BOOLEAN (BIT, NUMBER(1) or 1/0 where there is no boolean type), ISO date strings → DATE or TIMESTAMP when every value is a valid date, other strings → VARCHAR(longest) or TEXT, objects and arrays → JSON, JSONB, NVARCHAR(MAX) or CLOB. A column whose values mix kinds becomes text. Strings double their single quotes, and MySQL also doubles backslashes.

Example

The sample three customers become CREATE TABLE "customers" with id INTEGER, name VARCHAR(12), active BOOLEAN, balance NUMERIC(6,2), joined DATE, tags JSONB and address JSONB, then one INSERT with three rows. O'Brien is written 'O''Brien', Lin's missing tags and address and null balance become NULL, and Ada's address is stored as '{"city":"London"}'.