MCP 服务器

seedbase-mcp

Enables AI assistants to generate realistic, relationship-preserving test data from Seedbase projects into databases.

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 token option passed to the constructor.
The SEEDBASE_TOKEN environment variable.
The token field in ~/.seedbase/config.json (written by seedbase 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 });