koh-yoshimoto/mysql-mcp-server

MySQL MCP Server

A Model Context Protocol (MCP) server that provides MySQL database access through standardized tools.

Features

• Query Execution: Execute arbitrary SQL queries
• Schema Inspection: View table schemas
• Table Listing: List all tables in the database
• Query Analysis: Analyze query execution plans with optimization suggestions

Requirements

• Go 1.21+ (developed with Go 1.23.6)
• MySQL 5.7+ or MySQL 8.0+
• Make (for build commands)

Installation

Option 1: Download Pre-built Binary (Recommended)

Download the latest release for your platform:

Linux (amd64):

curl -L https://github.com/koh/mysql-mcp-server/releases/latest/download/mysql-mcp-server-linux-amd64.tar.gz | tar xz
chmod +x mysql-mcp-server
sudo mv mysql-mcp-server /usr/local/bin/

macOS (Apple Silicon):

curl -L https://github.com/koh/mysql-mcp-server/releases/latest/download/mysql-mcp-server-darwin-arm64.tar.gz | tar xz
chmod +x mysql-mcp-server
mv mysql-mcp-server /usr/local/bin/

macOS (Intel):

curl -L https://github.com/koh/mysql-mcp-server/releases/latest/download/mysql-mcp-server-darwin-amd64.tar.gz | tar xz
chmod +x mysql-mcp-server
mv mysql-mcp-server /usr/local/bin/

Windows:

# Download from https://github.com/koh/mysql-mcp-server/releases/latest
# Extract mysql-mcp-server-windows-amd64.zip
# Add to PATH or move mysql-mcp-server.exe to a directory in PATH

Option 2: Build from Source

1. Clone the repository:

git clone https://github.com/koh/mysql-mcp-server.git
cd mysql-mcp-server

2. Install dependencies and build:

make build
# Or use make setup for full development setup

Verify Installation

After installation, verify the server is accessible:

mysql-mcp-server --version

Configuration

The server uses environment variables for MySQL connection configuration:

• MYSQL_HOST: MySQL server host (default: localhost)
• MYSQL_PORT: MySQL server port (default: 3306)
• MYSQL_USER: MySQL username
• MYSQL_PASSWORD: MySQL password
• MYSQL_DATABASE: Database name to connect to

You can copy .env.example to .env and modify it with your credentials:

cp .env.example .env

Usage

With Claude Desktop

Add the server to your Claude Desktop configuration file:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "mysql": {
      "command": "mysql-mcp-server",
      "env": {
        "MYSQL_HOST": "localhost",
        "MYSQL_PORT": "3306",
        "MYSQL_USER": "your_user",
        "MYSQL_PASSWORD": "your_password",
        "MYSQL_DATABASE": "your_database"
      }
    }
  }
}

Direct Usage

Run the server directly:

export MYSQL_HOST=localhost
export MYSQL_PORT=3306
export MYSQL_USER=root
export MYSQL_PASSWORD=password
export MYSQL_DATABASE=testdb
./mysql-mcp-server

Available Tools

query

Execute SELECT queries to retrieve data from MySQL database. This tool is restricted to SELECT statements only for safety. Use the execute tool for data modification operations.

Parameters:

• query (required): SELECT statement only
• format (optional): Output format - json, table, csv, or markdown (default: table)

Example:

{
  "name": "query",
  "arguments": {
    "query": "SELECT * FROM users WHERE status = 'active' LIMIT 10",
    "format": "csv"
  }
}

execute

Execute INSERT, UPDATE, DELETE queries with safety checks. This tool implements a two-step execution process for safety:

1. First run with dry_run=true to preview affected rows
2. Then run with dry_run=false and the confirmation token to execute

Parameters:

• sql (required): INSERT, UPDATE, or DELETE statement
• dry_run (optional): If true, shows affected rows without executing (default: true)
• confirm_token (optional): Token from dry-run response, required when dry_run=false

Example - Step 1 (Dry Run):

{
  "name": "execute",
  "arguments": {
    "sql": "UPDATE users SET status = 'inactive' WHERE last_login < '2024-01-01'",
    "dry_run": true
  }
}

Response:

{
  "content": [
    {"type": "text", "text": "DRY RUN - Operation: UPDATE"},
    {"type": "text", "text": "This operation will affect 42 rows"},
    {"type": "text", "text": "To execute this query, run again with dry_run=false and the confirmation token below:"},
    {"type": "text", "text": "confirm_token: abc123def456"}
  ],
  "affected_rows": 42,
  "operation": "UPDATE",
  "confirm_token": "abc123def456"
}

Example - Step 2 (Execute):

{
  "name": "execute",
  "arguments": {
    "sql": "UPDATE users SET status = 'inactive' WHERE last_login < '2024-01-01'",
    "dry_run": false,
    "confirm_token": "abc123def456"
  }
}

schema

Get the schema of a MySQL table.

Parameters:

• table (required): The name of the table

Example:

{
  "name": "schema",
  "arguments": {
    "table": "users"
  }
}

tables

List all tables in the database.

Example:

{
  "name": "tables",
  "arguments": {}
}

explain

Analyze the execution plan of a MySQL query to understand performance. Supports both EXPLAIN and EXPLAIN ANALYZE.

Parameters:

• query (required): The SQL query to analyze
• analyze (optional): If true, runs EXPLAIN ANALYZE to get actual execution statistics (default: false)

Example - Basic EXPLAIN:

{
  "name": "explain",
  "arguments": {
    "query": "SELECT * FROM users WHERE email = 'test@example.com'"
  }
}

Example - EXPLAIN ANALYZE:

{
  "name": "explain",
  "arguments": {
    "query": "SELECT * FROM users WHERE age > 25",
    "analyze": true
  }
}

Note: EXPLAIN ANALYZE actually executes the query to gather real execution statistics, including actual row counts and timing information. Use with caution on queries that modify data or take a long time to execute.

Integration with AI Tools

VSCode Integration

Option 1: Using MCP Client Extensions

Configure in VSCode settings.json:

{
  "mcp.servers": {
    "mysql": {
      "command": "/path/to/mysql-mcp-server",
      "env": {
        "MYSQL_HOST": "localhost",
        "MYSQL_USER": "root",
        "MYSQL_PASSWORD": "password",
        "MYSQL_DATABASE": "mydb"
      }
    }
  }
}

Option 2: Custom VSCode Extension

Create a custom extension that spawns the MCP server. See for implementation details.

Cursor Integration

Cursor supports MCP servers through its configuration:

1. Open Cursor Settings
2. Navigate to "AI" → "Model Context Protocol"
3. Add server configuration:

{
  "mysql": {
    "command": "/path/to/mysql-mcp-server",
    "env": {
      "MYSQL_HOST": "localhost",
      "MYSQL_USER": "root",
      "MYSQL_PASSWORD": "password",
      "MYSQL_DATABASE": "mydb"
    }
  }
}

GitHub Copilot Integration

GitHub Copilot doesn't directly support MCP servers, but you can create a bridge through VSCode extensions. See for detailed implementation.

Generic Integration Pattern

For any tool that supports subprocess communication:

const { spawn } = require('child_process');

class MCPClient {
    constructor(serverPath, env) {
        this.server = spawn(serverPath, [], { env });
        // ... handle communication
    }

    async callTool(name, arguments) {
        return this.request('tools/call', { name, arguments });
    }
}

// Usage
const client = new MCPClient('/path/to/mysql-mcp-server', {
    MYSQL_HOST: 'localhost',
    MYSQL_USER: 'root',
    MYSQL_PASSWORD: 'password',
    MYSQL_DATABASE: 'mydb'
});

For complete integration examples and troubleshooting, see .

Testing

For detailed testing instructions, see .

Quick start:

# Setup and run interactive test client
make setup
make test-client

Development

Run tests:

make test

Run the server:

make run

Security Considerations

• The query tool is restricted to SELECT statements only to prevent accidental data modification
• The execute tool requires a two-step confirmation process for all data modification operations
• Never expose this server to untrusted clients
• Use appropriate MySQL user permissions
• Consider using read-only database users when possible
• The dry-run feature allows you to preview the impact of UPDATE/DELETE operations before execution
• Confirmation tokens expire after 5 minutes for security
• Keep your database credentials secure

License

MIT