Graphjin
Connect AI to Your Databases via MCP
npx graphjin serve LLMs struggle with databases, diverse SQL dialects, custom query methods like MongoDB, complex joins. GraphJin is a compiler that auto-learns your entire schema and generates efficient queries from simple GraphQL. Even smaller models can describe exactly what data they need, and GraphJin handles the rest. Nothing like this exists for AI, works with Claude Desktop or any MCP client.
Works with your database
Quick Start
npx graphjin serveUse with Node.js projects
Works on macOS, Windows, and Linux. Supports PostgreSQL, MySQL, SQLite, MongoDB, and more.
Simple & Powerful
Write GraphQL, get optimized SQL. Queries, mutations, and aggregations — all in one query.
query {
products(limit: 10, order_by: { price: desc }) {
id
name
price
owner {
full_name
email
}
}
}{
"products": [
{
"id": 104,
"name": "Mechanical Keyboard",
"price": 149.99,
"owner": {
"full_name": "Alice Dev",
"email": "alice@example.com"
}
}
]
}Why GraphJin?
Built for the AI era. No resolvers, no ORMs, no N+1 problems.
AI Native
Works with Claude Desktop, GPT, and any MCP-compatible AI. The AI describes what data it needs in GraphQL — GraphJin handles the SQL.
Auto Discovery
Point it at your database. Tables, columns, relationships, and constraints are introspected automatically.
Single Query
Nested queries across 10 tables? Still compiles to one SQL statement. N+1 is eliminated at the compiler level.
Multi-Database
Postgres, MySQL, MongoDB, SQLite, MSSQL, Oracle, YugabyteDB, CockroachDB. Same GraphQL, any backend.
Production Ready
Role-based access control, JWT authentication, query allow-lists, and row-level security built in.
No Resolvers
Skip the boilerplate. No resolver functions, no type definitions, no ORM mappings. Your schema is the API.