QueryMind AI v2.0

An agentic AI system for natural language database interactions, built on the Model Context Protocol (MCP).

Overview

QueryMind AI is an autonomous AI agent that translates natural language into database operations. Unlike traditional chatbots, this agentic system reasons, plans, and executes multi-step database tasks independently—deciding which tools to invoke, interpreting results, and iterating until the user's intent is fulfilled.

Ask questions like "Show me all tables" or "Find users created last week" and watch the agent autonomously query your database and return formatted results.

Agentic Capabilities

• Autonomous Decision-Making: The AI agent analyzes user intent and selects appropriate database tools without explicit instructions
• Tool Orchestration: Dynamically chains MCP tools (list tables → describe schema → execute query) based on context
• Self-Correcting: Interprets tool outputs and adjusts strategy when needed
• Context-Aware: Maintains conversation context to handle follow-up queries intelligently

Architecture

┌────────┐         ┌──────────────┐         ┌─────────────┐
│  User  │────────▶│ React Client │────────▶│  AI Agent   │
└────────┘  Query  └──────┬───────┘  Intent └──────┬──────┘
                          │                        │
                          │   Tool Calls           │ Reasoning
                          │◀───────────────────────┘
                          │
                          │ Execute Tools
                          ▼
                   ┌─────────────┐
                   │  MCP Server │
                   └──────┬──────┘
                          │
                          │ SQL Query
                          ▼
                   ┌─────────────┐
                   │  Database   │
                   └──────┬──────┘
                          │
                          │ Results
                          ▼
┌────────┐         ┌──────────────┐         ┌─────────────┐
│  User  │◀────────│ React Client │────────▶│  AI Agent   │
└────────┘Response └──────────────┘ Format  └─────────────┘

• Client: React + Vite frontend that orchestrates the agentic loop
• AI Agent: Groq-powered Llama 3.3 70B for reasoning, tool selection, and response synthesis
• MCP Server: Exposes database tools via SSE transport following the Model Context Protocol
• Database: SQL database (MySQL, PostgreSQL, etc.)

Features

• Natural language to SQL conversion
• Real-time SSE connection for responsive interactions
• Database schema exploration (list tables, describe structure)
• Execute SELECT queries safely
• Execute INSERT/UPDATE/DELETE/CREATE/DROP operations
• Markdown-formatted results with tables

Quick Start

Prerequisites

• Node.js 18+
• MySQL database
• Groq API key

1. Clone the repository

git clone https://github.com/athulya-anil/QueryMindAI.git
cd QueryMindAI

2. Setup the Server

cd mcp-server
npm install
npm run build
npm start

3. Connect Your Database

Create a .env file in mcp-server/:

MYSQL_HOST=localhost
MYSQL_PORT=3306
MYSQL_USER=your_username
MYSQL_PASSWORD=your_password
MYSQL_DATABASE=your_database
MYSQL_CONN_LIMIT=10

Replace the values with your database credentials.

4. Setup the Client

cd mcp-client
npm install
npm run dev

Create a .env file in mcp-client/:

VITE_GROQ_API_KEY=your_groq_api_key

5. Open in Browser

Navigate to http://localhost:5173 and start querying!

Available MCP Tools

Tool	Description
list_tables	List all tables in the database
describe_table	Get schema information for a table
execute_query	Execute SELECT queries
execute_update	Execute INSERT/UPDATE/DELETE/CREATE/DROP queries

Tech Stack

• Frontend: React 19, Vite, TypeScript, react-markdown
• Backend: Express 5, TypeScript, mysql2
• Protocol: Model Context Protocol (MCP) with SSE transport
• AI Engine: Groq API with Llama 3.3 70B Versatile

License

MIT © 2025 Athulya Anil