SeedBase — Synthetic Test Data
RemoteAbout SeedBase — Synthetic Test Data
SeedBase provides schema-aware, reproducible synthetic test data for database projects. Its MCP server lets AI assistants import schemas, inspect DDL, and generate foreign-key-consistent datasets as SQL.
Key features
- 01List and create projects
- 02Import SQL, CSV, JSON, or ORM schemas
- 03Retrieve schema DDL by dialect
- 04Generate seeded, FK-consistent SQL data
- 05Write large results to a local file
Use cases
- 01Prepare realistic data for database testing
- 02Import an existing schema to SeedBase
- 03Generate repeatable datasets for development or CI
MCP support
The package includes an MCP server with five tools for listing and creating projects, importing schemas, retrieving DDL, and generating test data. It can run locally over stdio or connect through the hosted Streamable-HTTP endpoint.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://seedba.se/mcpclaude mcp add --transport http seedbase-node https://seedba.se/mcp{"mcpServers":{"seedbase-node":{"type":"http","url":"https://seedba.se/mcp"}}}Open in Cursor{"servers":{"seedbase-node":{"type":"http","url":"https://seedba.se/mcp"}}}Open in VS Code{"mcpServers":{"seedbase-node":{"type":"http","url":"https://seedba.se/mcp"}}}{"mcpServers":{"seedbase-node":{"serverUrl":"https://seedba.se/mcp"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
https://seedba.se/mcpAuth requiredTools
6This tool takes no parameters.
Details
From the project README
@seedbase/client
Generate realistic, relationship-preserving, privacy-safe test data for your databases — and pull it straight into your local or CI database.
Seedbase lives on seedbase.dev: you model (or import) a schema there, generate datasets, and use this package to pull them into Postgres, MySQL, SQLite and more. Schema-aware, foreign-key-correct, reproducible by seed.
This is the Node.js client, a counterpart to the Python SDK.
Install
npm install @seedbase/client
Zero runtime dependencies — pure ESM, built on the native fetch of Node 18+.
Quickstart
import { SeedbaseClient } from "@seedbase/client";
// Token from the argument, $SEEDBASE_TOKEN, or ~/.seedbase/config.json
const client = new SeedbaseClient({ token: "dr_sk_..." });
// Trigger a generation and wait for it to finish
const gen = await client.generate(projectId, { seed: 42, wait: true });
// Download the result (Uint8Array)
const bytes = await client.download(gen.id, { format: "sql" });
import { writeFile } from "node:fs/promises";
await writeFile("dump.sql", bytes);
MCP server (Claude Code, Claude Desktop & friends)
This package ships seedbase-mcp — a zero-dependency Model Context Protocol
server that lets AI assistants generate test data for you. Describe what you
need ("fill my Shop project with MySQL test data") and the assistant drives
SeedBase end-to-end through five tools:
| Tool | What it does |
|---|---|
list_projects | List your SeedBase projects (id, name, database type) |
create_project | Create a new, empty project |
import_schema | Import a schema from SQL DDL (raw pg_dump --schema-only works), CSV/JSON or ORM model code |
get_ddl | Get a project's schema as CREATE TABLE statements, per dialect |
generate_test_data | Generate a fresh FK-consistent dataset and return it as SQL (large results are written to a local file, never truncated) |
Hosted (zero install) — point any Streamable-HTTP MCP client at
https://seedbase.dev/mcp with an Authorization: Bearer dr_sk_... header:
claude mcp add-json seedbase '{"type":"http","url":"https://seedbase.dev/mcp","headers":{"Authorization":"Bearer dr_sk_..."}}'
Local via Claude Code (stdio):
claude mcp add-json seedbase '{"type":"stdio","command":"npx","args":["-y","-p","@seedbase/client","seedbase-mcp"],"env":{"SEEDBASE_API_KEY":"dr_sk_..."}}'
Claude Desktop (claude_desktop_config.json):
{
"mcpServers": {
"seedbase": {
"command": "npx",
"args": ["-y", "-p", "@seedbase/client", "seedbase-mcp"],
"env": { "SEEDBASE_API_KEY": "dr_sk_..." }
}
}
}
Create a free account at seedbase.dev/register (no
credit card), then create an API key under Settings → API keys. The free tier is
enough to generate full, foreign-key-consistent datasets. The server is stdio-only,
talks exclusively to https://seedbase.dev, and stores nothing locally.
Authentication
The token is resolved in this order:
- The
tokenoption passed to the constructor. - The
SEEDBASE_TOKENenvironment variable. - The
tokenfield in~/.seedbase/config.json(written byseedbase login).
API keys with the dr_sk_ prefix are sent as Authorization: Bearer ..., other
tokens as Authorization: Token .... Get a key at
seedbase.dev/settings?tab=api-keys.
API
new SeedbaseClient({
token, // optional, see resolution order above
apiUrl, // default "https://seedbase.dev/api/v1" (https enforced, http only for localhost)
configPath, // override ~/.seedbase/config.json
requestTimeout, // per-request timeout in ms, default 30000
fetch, // inject a custom fetch (e.g. for tests)
});
| Method | Description |
|---|---|
listProjects() | All datasets/projects (paginated, followed automatically). |
getProject(projectId) | A single project. |
listGenerations(projectId) | Generations for a project (paginated). |
getGeneration(generationId) | A single generation. |
generate(projectId, opts) | Trigger a generation. opts: { seed, rows, format, rebaseTo, wait, timeout, pollInterval }. With wait: true it polls until the generation reaches completed/failed/cancelled. |
download(generationId, { format }) | Download the generated artifact as a Uint8Array. format defaults to "sql". |
seededRows(projectId, { seed, rows }) | Generate and return the rows as { tableName: [row, ...] }, in foreign-key-safe order. |
exportConfig(projectId) | The project's engine config as an object. |
importConfig(projectId, config) | Replace the project's engine config. |
All methods are async and return Promises. Failures throw a SeedbaseError
(with .statusCode for HTTP errors), carrying a readable message that includes
the server's detail or field errors.
import { SeedbaseError } from "@seedbase/client";
try {
await client.getProject("missing");
} catch (err) {
if (err instanceof SeedbaseError) {
console.error(err.statusCode, err.message);
}
}
Prisma seed
Fill a Prisma-managed database with realistic, foreign-key-consistent data, in
one call. Your schema must already exist (your prisma migrate owns it);
SeedBase only fills it. Free tier.
// prisma/seed.ts
import { PrismaClient } from "@prisma/client";
import { SeedbaseClient } from "@seedbase/client";
import { seedPrisma } from "@seedbase/client/prisma";
const prisma = new PrismaClient();
const client = new SeedbaseClient({ token: process.env.SEEDBASE_TOKEN });
await seedPrisma(prisma, client, { project: process.env.SEEDBASE_PROJECT, seed: 42 });
Then run prisma db seed. A runnable demo (offline, no account) is in
examples/prisma-seed-demo.mjs.
Links
- Website: https://seedbase.dev
- Docs: https://seedbase.dev/docs
- API keys: https://seedbase.dev/settings?tab=api-keys
MIT licensed.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.