Qora

Talk to your data. Get clear answers, insights, and analysis.

Qora lets AI assistants like Cursor and Claude safely explore your database and answer questions about your data in plain English.

Ask things like:

"How many orders did we receive last month?"

"Which products are selling the most?"

"Show me the tables related to customers."

Qora looks up the answers directly from your database and gives them back to your AI assistant.

Qora is read-only by design. It can inspect and query your data, but it cannot insert, update, or delete anything.

How it works

flowchart LR
  You["You ask a question"] --> AI["AI assistant<br/>(Cursor, Claude, …)"]
  AI -->|"uses Qora tools"| Qora["Qora"]
  Qora -->|"explore & read-only queries"| DB[("Your database")]
  DB -->|"rows & schema info"| Qora
  Qora -->|"safe results"| AI
  AI -->|"clear answer"| You

1. You ask something.

2. Your AI assistant calls Qora.

3. Qora looks up schemas/tables or runs a read-only query.

4. Results go back to the assistant, which answers you.

Related MCP server: GraphJin

What can Qora do?

• Explore your data: See what schemas and tables are available in your database.

• Understand your database: Inspect tables and their columns without opening a database client.

• Ask questions: Ask things like "How many customers signed up this month?" or "What's the average order value?"

• Analyze your data: Find trends, compare numbers, and understand what's happening in your business.

Read-only by design

There are multiple layers of protection:

1. Read-only queries: Qora only allows queries that retrieve data.

2. Read-only database user: Connect Qora using a database account that cannot modify data.

3. Result limits: Responses are limited in size to prevent accidentally returning huge amounts of data.

4. Host restrictions: Web access can be restricted to specific hosts.

Even if an AI assistant makes a mistake, Qora is designed to prevent it from modifying your database.

Getting started

Requirements

• Node.js

• pnpm

• PostgreSQL

Create a PostgreSQL user with read-only access and add its connection string to .env.

pnpm install
cp .env.example .env

Then configure:

DATABASE_URL=postgresql://readonly_user:password@localhost:5432/prod

Try it with sample data

A sample database is included in seed.sql.

Create a database called prod, then load the sample data:

psql -d prod -f seed.sql

Start Qora

Local development

pnpm dev

Connect directly from Cursor or Claude

TRANSPORT=stdio pnpm start

Production

pnpm build
pnpm start

Connect Qora to Cursor or Claude Desktop

Qora uses the Model Context Protocol (MCP), an open standard that allows AI assistants to securely use external tools and data.

Add Qora to your MCP configuration:

{
  "mcpServers": {
    "qora": {
      "command": "node",
      "args": ["--env-file=.env", "./node_modules/tsx/dist/cli.mjs", "src/server.ts"],
      "cwd": "/absolute/path/to/qora",
      "env": {
        "TRANSPORT": "stdio"
      }
    }
  }
}

Once connected, your AI assistant can use Qora to explore and query your database.

Use Qora over the web

Qora can also run as an HTTP-based MCP server.

Start Qora:

TRANSPORT=http pnpm start

Your MCP endpoint will be available at:

http://your-host:8080/mcp

For a quick private tunnel during development:

ngrok http 8080

Important: Don't expose your database or Qora endpoint publicly without appropriate authentication and network controls.

What Qora gives your AI assistant

The AI can combine these tools to first understand your database and then find the information it needs.

Configuration

Copy .env.example and configure:

Development