SynthForge is a web app for generating realistic synthetic and test data across related database tables in a single pass: every foreign key points at a parent row that actually exists.
Design a schema once from plain English (AI), a visual editor, pasted SQL DDL, CSV assist, or templates. Then generate rows for every table together: 152 field types across 19 categories, statistical distributions, conditional column values, per-relationship cardinality, and dialect-aware exports (PostgreSQL, MySQL, SQLite, SQL Server, MariaDB, DuckDB, CockroachDB) plus CSV, JSON, JSONL, and Parquet. Large flat exports can stream out-of-core on the server. ML training-dataset templates are included for common use cases.
Automate from scripts, CI, or AI agents with personal API keys, a public OpenAPI at api.synthforge.io, a TypeScript SDK on npm (@synthforgeio/sdk), and a hosted MCP server for agent hosts such as Claude and Cursor.
Pricing: the first 20,000 rows of every dataset are free in the web app, and AI schema design is free there at any size. Past that, generation uses credits (a credit is one cent), sold in packs from $5 with no subscription. API keys are metered from the first row, and your first key comes with 250 credits. An account is required; no credit card to sign up.
SynthForge generates greenfield data from a schema. It does not de-identify production databases or provide differential privacy.