Connect RushDB

Start with the path that matches how you work. Every path below works with RushDB Cloud (sign in at app.rushdb.com and use an API key) or with a self-hosted RushDB engine (the rushdb binary on your own machines, reached with its URL and token).

Goal	Fastest path	Cloud	Self-hosted
Build an application or data pipeline	Use an SDK	✓	✓
Give Cursor, Claude Desktop, or VS Code RushDB tools	Local MCP server	✓	✓
Use RushDB from ChatGPT or Claude.ai	Hosted MCP connector	✓	—
Teach an AI agent correct RushDB workflows	Install Agent Skills	✓	✓
Add persistent memory to an agent	Agent memory	✓	✓
Query with PostgreSQL or Neo4j drivers, or gRPC	SQL and Cypher clients	—	✓

---

Use an SDK

The TypeScript and Python SDKs have one API for both. On RushDB Cloud, pass an API key; for a self-hosted engine, pass its token with direct: true and its URL.

import RushDB from '@rushdb/javascript-sdk'

const db = new RushDB(process.env.RUSHDB_API_KEY!) // RushDB Cloud
// const db = new RushDB(process.env.RUSHDB_TOKEN!, { direct: true, engineUrl: 'http://127.0.0.1:3000' })

await db.records.importJson({
  label: 'PROJECT',
  data: {
    name: 'RushDB adoption',
    TASK: [
      { title: 'Connect RushDB', status: 'done' },
      { title: 'Query linked records', status: 'pending' }
    ]
  }
})

One nested import creates the records and the relationships between them.

Connect with TypeScript or Python →

---

Local MCP Server

For Claude Desktop, Cursor, VS Code, and other local MCP clients, run the MCP server with a RushDB Cloud API key:

{
  "mcpServers": {
    "rushdb": {
      "command": "npx",
      "args": ["-y", "@rushdb/mcp-server"],
      "env": {
        "RUSHDB_API_KEY": "your-api-key-here"
      }
    }
  }
}

or point it at your own engine with RUSHDB_MODE=direct, RUSHDB_ENGINE_URL and RUSHDB_ENGINE_TOKEN. Restart your client, then ask:

Call getSchemaMarkdown and show me what labels exist in my RushDB project.

Choose your MCP client and mode →

---

Hosted MCP Connector

For ChatGPT, Claude.ai, and other web-based MCP clients, add this URL as a connector:

https://mcp.rushdb.com/mcp

Sign in with your RushDB account and choose the project to expose. No local installation or API key is required. The hosted connector is part of RushDB Cloud.

Set up a hosted MCP connector →

---

Install Agent Skills

Install RushDB Agent Skills so your assistant knows how to query, model data, and manage persistent memory:

npx skills add rush-db/skills

Start a new agent session after installation so the skills are discovered.

Install and verify Agent Skills →

---

Agent Memory

Give an agent durable, searchable memory through MCP, or natively in OpenClaw and Hermes:

• Agent Memory Quickstart — MCP plus skills, any agent
• OpenClaw Native Memory and Hermes Native Memory Provider — automatic recall and capture
• Agents — the shared memory contract behind them

On a self-hosted engine, recall by meaning needs an embedding model of your own; each integration explains how to supply one.

---

Engine Protocols

A self-hosted engine also serves its own interfaces, next to the SDKs. SQL and Cypher clients shows how to turn them on and connect psql, node-postgres, psycopg and the Neo4j drivers.

Interface	Use it for
Native HTTP protocol	The complete engine API, from any language
PostgreSQL wire	SQL from psql, drivers and BI tools
Bolt	openCypher from Neo4j drivers
gRPC	The native protocol over gRPC
CLI	rushdb up and operational commands

---

Choose What to Build

Use case	Guide
Persistent agent memory	Agent Memory Quickstart
Keyword and hybrid search	Keyword Search
Semantic search over your data	Semantic Search
Natural-language search box [Cloud]	Smart Search
Graph-enriched retrieval	GraphRAG Tutorial
Explore an unknown dataset safely	Discovery Queries