JSON → SQL INSERT Generator

JSON to SQL INSERT generator

Paste a JSON object or an array of row objects and get SQL INSERT statements you can run against your database. Everything happens locally in WebAssembly — your data is never uploaded, so it's safe for real records.

How it works

• Input — a single JSON object becomes one row; an array of objects becomes one row each. The columns are the union of every row's keys, in the order they first appear (turn on Sort columns alphabetically to reorder them). A key that's missing from a given row is treated as null, so heterogeneous rows still line up under one column list.
• Dialect — pick MySQL (`backticks`, 1/0 booleans, backslash escaping), PostgreSQL or ANSI ("double quotes", TRUE/FALSE), SQLite ("double quotes", 1/0), or SQL Server ([brackets], 1/0). The dialect also drives placeholder syntax and inferred column types.
• Values — literal inlines properly-escaped SQL literals (a ' becomes ''; MySQL also doubles a \). Placeholder emits a parameterized / prepared statement (? for MySQL, SQLite and SQL Server as @pN, $1, $2… for Postgres) and lists the bound values in a trailing -- params: comment.
• Batch vs per-row — Batch produces one INSERT … VALUES (…),(…) for all rows; turn it off for a separate INSERT per row.
• CREATE TABLE — optionally emit a CREATE TABLE first, with each column's type inferred from the data (integer, float, boolean, or text). Name a primary key column, and optionally prepend DROP TABLE IF EXISTS.
• Nulls — choose whether a JSON null (or a missing key) becomes NULL, the keyword DEFAULT (so the column's own default applies), or an empty string ''.
• Identifiers — table and column names are quoted for the dialect by default. Turn quoting off to emit bare identifiers; unsafe names (spaces, punctuation, a leading digit) are rejected with a clear error instead of producing broken or injectable SQL. A schema-qualified table like public.users is quoted part-by-part ("public"."users").

Example

Input (Postgres, with CREATE TABLE and id as the primary key):

[{"id": 1, "email": "[email protected]", "active": true},
 {"id": 2, "email": "[email protected]", "active": false}]

Output:

CREATE TABLE "accounts" (
  "id" INTEGER,
  "email" TEXT,
  "active" BOOLEAN,
  PRIMARY KEY ("id")
);
INSERT INTO "accounts" ("id", "email", "active") VALUES
(1, '[email protected]', TRUE),
(2, '[email protected]', FALSE);

Limits & edge cases

• Input must be a JSON object or an array of objects. A top-level scalar, an array of scalars ([1, 2, 3]), or an empty array is rejected with a message saying what was expected.
• Nested objects/arrays in a cell are stored as a compact JSON string literal (e.g. {"k":[1,2]} → '{"k":[1,2]}') so nothing is dropped — the tool does not flatten them into extra columns or normalize into related tables.
• CREATE TABLE types are inferred only as integer / float / boolean / text; no lengths, precision, indexes, or foreign keys are generated.
• Numbers are emitted exactly as they appear in the JSON. Everything runs in your browser, so very large inputs are bounded by your device's memory.

FAQ