MySQL MCP Server

A high-quality Model Context Protocol (MCP) server implementation for MySQL databases. This server enables AI assistants like Claude to interact with MySQL databases through a standardized protocol.

Version: 0.2.0 | Protocol: MCP 2025-03-26 | Rust: 1.70+ | Status: Production Ready

Table of Contents

• Features
• Installation
• Quick Start (5 Minutes)
• Usage
• Available Tools
• Database Context Feature
• Security Considerations
• Architecture
• Troubleshooting
• Development
• Deployment Guide
• Contributing
• License
• Support

Features

• Schema Inspection: Retrieve table schemas and structure information
• Query Execution: Execute SQL queries (read-only by default for safety)
• Data Manipulation: Insert, update, and delete operations
• Database Context: Specify which database to use per query
• Safety Controls: Configurable query restrictions to prevent dangerous operations
• Connection Management: Robust connection handling with retry logic and pooling
• Error Handling: Comprehensive error reporting with detailed messages
• JSON-RPC 2.0 Protocol: Standardized communication via stdio

Installation

Prerequisites

• Rust 1.70+
• MySQL 5.7+ or MariaDB 10.2+
• Access to a MySQL database

Building from Source

git clone <repository-url>
cd mcp-server-mysql
cargo build --release

The compiled binary will be available at target/release/mcp-server-mysql.

From Release Package

# Extract the package
tar -xzf mcp-server-mysql-v0.2.0-linux-x86_64.tar.gz

# Move binary to system path (optional)
sudo cp mcp-server-mysql /usr/local/bin/

# Verify installation
mcp-server-mysql --version

Quick Start (5 Minutes)

Step 1: Build the Server

cargo build --release

The binary will be at target/release/mcp-server-mysql

Step 2: Test the Connection

./target/release/mcp-server-mysql \
  --host localhost \
  --username root \
  --password yourpassword \
  --database testdb

You should see: "MCP MySQL Server started and ready to accept connections"

Step 3: Configure Claude Desktop

Edit your Claude Desktop configuration file:

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

Add this configuration:

{
  "mcpServers": {
    "mysql": {
      "command": "/absolute/path/to/mcp-server-mysql",
      "args": [
        "--host", "localhost",
        "--port", "3306",
        "--username", "your_username",
        "--password", "your_password",
        "--database", "your_database"
      ]
    }
  }
}

Security Note: For production use, consider using environment variables or a secure secrets management solution instead of hardcoding passwords in the configuration file.

Step 4: Restart Claude Desktop

Close and reopen Claude Desktop completely. You should see a small hammer icon indicating the MCP server is connected.

Step 5: Try it Out!

Ask Claude:

• "Can you show me the schema for the users table in my MySQL database?"
• "Query the database and show me the first 10 rows from the products table"
• "What tables are in my database?"

Usage

Command Line Arguments

mcp-server-mysql \
  --host localhost \
  --port 3306 \
  --username your_username \
  --password your_password \
  --database your_database \
  --allow-dangerous-queries false

Arguments Reference

Configuration with Claude Desktop

Add this configuration to your Claude Desktop config file:

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

{
  "mcpServers": {
    "mysql": {
      "command": "/path/to/mcp-server-mysql",
      "args": [
        "--host", "localhost",
        "--port", "3306",
        "--username", "your_username",
        "--password", "your_password",
        "--database", "your_database"
      ]
    }
  }
}

Available Tools

1. mysql (Schema Inspection)

Retrieve database schema information for tables.

Parameters:

• table_name (string): Name of the table to inspect, or "all-tables" to get all table schemas

Example:

{
  "table_name": "users"
}

Returns:

• Column information (name, type, nullable, defaults, keys)
• Index information
• Table constraints

2. query (SQL Execution)

Execute SQL queries on the database.

Parameters:

• query (string): SQL query to execute
• database (string, optional): Database name to use for this specific query

Example:

{
  "query": "SELECT * FROM users WHERE active = 1 LIMIT 10",
  "database": "my_database"
}

Safety:

• By default, only SELECT queries are allowed
• Use --allow-dangerous-queries flag to enable INSERT/UPDATE/DELETE
• Dangerous keywords are blocked unless explicitly enabled

3. insert (Insert Data)

Insert data into a specified table.

Parameters:

• table_name (string): Name of the table
• data (object): Key-value pairs of column names and values

Example:

{
  "table_name": "users",
  "data": {
    "username": "john_doe",
    "email": "john@example.com",
    "active": true
  }
}

Returns: Last insert ID

4. update (Update Data)

Update data in a specified table based on conditions.

Parameters:

• table_name (string): Name of the table
• data (object): Key-value pairs of columns to update
• conditions (object): Key-value pairs for WHERE clause

Example:

{
  "table_name": "users",
  "data": {
    "email": "newemail@example.com",
    "updated_at": "2024-01-15 10:30:00"
  },
  "conditions": {
    "id": 123
  }
}

Returns: Number of affected rows

5. delete (Delete Data)

Delete data from a specified table based on conditions.

Parameters:

• table_name (string): Name of the table
• conditions (object): Key-value pairs for WHERE clause

Example:

{
  "table_name": "users",
  "conditions": {
    "id": 123
  }
}

Returns: Number of affected rows

Warning: Always specify conditions to avoid deleting all rows!

Database Context Feature

The Problem

Previously, database context was not maintained between queries:

-- Query 1
USE dev_database;  -- Succeeds

-- Query 2 (new connection from pool)
SELECT * FROM my_table;  -- ❌ Fails: context was lost

The Solution

Use the optional database parameter on each query:

{
  "query": "SELECT * FROM my_table",
  "database": "dev_database"
}

Benefits

1. Explicit and Clear: Know exactly which database each query uses
2. No Hidden State: Each query is independent
3. Backward Compatible: Existing queries without parameter still work
4. No Race Conditions: Each query gets its own connection
5. Simple to Use: Just add "database": "name" to query arguments

Usage Examples

Basic Query with Database Parameter

{
  "query": "SELECT * FROM crm_sites LIMIT 10",
  "database": "dev_smartConnect_za"
}

Query Without Database Parameter (Uses Default)

{
  "query": "SELECT * FROM users WHERE active = 1"
}

Uses the database specified in --database startup argument.

Multiple Databases in Same Session

// Query database 1
{
  "query": "SELECT COUNT(*) FROM customers",
  "database": "production_db"
}

// Query database 2
{
  "query": "SELECT COUNT(*) FROM test_data",
  "database": "test_db"
}

Before vs After

Before (Required fully qualified names):

SELECT * FROM dev_smartConnect_za.crm_sites
  JOIN dev_smartConnect_za.crm_orgs ON ...
WHERE dev_smartConnect_za.crm_sites.active = 1;

After (Clean and simple):

{
  "query": "SELECT * FROM crm_sites JOIN crm_orgs ON ... WHERE active = 1",
  "database": "dev_smartConnect_za"
}

Common Scenarios

Scenario 1: Single Database Project

Set default database and omit the parameter:

# Startup
--database my_project_db

# Query (no database parameter needed)
{
  "query": "SELECT * FROM users"
}

Scenario 2: Multiple Database Project

Specify database for each query:

// Customer database
{ "query": "...", "database": "customers_db" }

// Orders database
{ "query": "...", "database": "orders_db" }

// Analytics database
{ "query": "...", "database": "analytics_db" }

Error Handling

Error Code -32005: Connection Acquisition Failed

Cause: Connection pool exhausted
Solution: Retry after a moment

Error Code -32006: Database Context Switch Failed

Cause: Database doesn't exist or user lacks permissions
Solution: Verify database exists and user has access

Best Practices

✅ DO

• Specify database explicitly for production queries
• Use descriptive database names in your queries
• Test with SELECT DATABASE() to verify context
• Group queries by database for clarity

❌ DON'T

• Mix qualified and unqualified names in the same query
• Assume persistence - specify database for each query
• Use special characters in database names if possible
• Forget to verify user permissions for all databases

Security Considerations

Read-Only Mode (Default)

By default, the server operates in read-only mode, allowing only SELECT queries. This prevents accidental data modification or deletion.

Dangerous Queries Mode

Enable write operations with --allow-dangerous-queries:

mcp-server-mysql --username user --password pass --database mydb --allow-dangerous-queries true

Use with caution! This enables:

• INSERT statements
• UPDATE statements
• DELETE statements
• Other potentially destructive operations

SQL Injection Protection

• Table names are validated to contain only alphanumeric characters and underscores
• All data values are parameterized using prepared statements
• Database names are escaped by replacing backticks with double backticks
• No raw SQL concatenation is performed

Connection Security

• Supports standard MySQL SSL/TLS connections
• Connection strings can be configured securely
• Passwords can be provided via environment variables
• Consider using dedicated database users with limited permissions

Production Deployment Security

1. Use dedicated database user:

   sqlCopy

   CREATE USER 'mcp_user'@'localhost' IDENTIFIED BY 'secure_password';
   GRANT SELECT ON your_database.* TO 'mcp_user'@'localhost';
   FLUSH PRIVILEGES;
   

2. Enable write access only when needed:

   bashCopy

   --allow-dangerous-queries true  # Use with caution!
   

3. Use environment variables (future enhancement): Consider wrapping the binary in a shell script that reads from env vars.

Architecture

System Overview

┌─────────────────────────────────────────────────────┐
│           MCP Client (e.g., Claude)                 │
│  Sends: {query, database}                           │
└────────────────────────┬────────────────────────────┘
                         │ JSON-RPC 2.0 (stdio)
                         ▼
┌─────────────────────────────────────────────────────┐
│      MCP MySQL Server (Rust)                        │
│                                                      │
│  execute_query(query, database, pool)               │
│  ├─ If database param:                              │
│  │  ├─ Acquire connection from pool                 │
│  │  ├─ Execute: USE `database`                      │
│  │  └─ Execute: [user's query]                      │
│  └─ Else:                                           │
│     └─ Execute query on pool (default database)     │
└────────────────────────┬────────────────────────────┘
                         │
                         ▼
┌─────────────────────────────────────────────────────┐
│      MySQL Connection Pool (5 connections)          │
└────────────────────────┬────────────────────────────┘
                         │
                         ▼
┌─────────────────────────────────────────────────────┐
│      MySQL/MariaDB Server                           │
└─────────────────────────────────────────────────────┘

Sequence: Query with Database Parameter

Client          MCP Server       Connection Pool      MySQL Server
  │                 │                    │                  │
  │  query +        │                    │                  │
  │  database       │                    │                  │
  ├────────────────>│                    │                  │
  │                 │                    │                  │
  │                 │ acquire()          │                  │
  │                 ├───────────────────>│                  │
  │                 │ <connection>       │                  │
  │                 │<───────────────────┤                  │
  │                 │                    │                  │
  │                 │ USE database       │                  │
  │                 ├────────────────────┼─────────────────>│
  │                 │ OK                 │                  │
  │                 │<────────────────────┼──────────────────┤
  │                 │                    │                  │
  │                 │ SELECT query       │                  │
  │                 ├────────────────────┼─────────────────>│
  │                 │ Results            │                  │
  │                 │<────────────────────┼──────────────────┤
  │                 │                    │                  │
  │                 │ release()          │                  │
  │                 ├───────────────────>│                  │
  │  Results        │                    │                  │
  │<────────────────┤                    │                  │

Connection Pool Management

Pool (5 connections)
┌────┐ ┌────┐ ┌────┐ ┌────┐ ┌────┐
│ C1 │ │ C2 │ │ C3 │ │ C4 │ │ C5 │
└────┘ └────┘ └────┘ └────┘ └────┘

Key Properties:
• Each query gets its own connection instance
• Database context is set per connection, per query
• No state persists between queries
• Fully thread-safe and concurrent

Technical Details

• Protocol Version: MCP 2025-03-26
• Transport: stdio (JSON-RPC 2.0)
• Connection Pooling: Max 5 connections
• Retry Logic: Automatic reconnection on transient failures
• Performance Overhead: ~50-200 microseconds per query with database parameter

Troubleshooting

Connection Failures

If you encounter connection errors:

1. Check MySQL is running:

   bashCopy

   mysql -h localhost -u your_username -p
   

2. Verify credentials:

   • Ensure the username and password are correct
   • Confirm the user has access to the specified database

3. Check network access:

   • Verify the host and port are correct
   • Ensure no firewall is blocking the connection

4. Review server logs:

   • The server logs to stderr
   • Check for detailed error messages

Common Errors

"Database connection failed"

• MySQL server may not be running
• Incorrect host/port configuration
• Network connectivity issues

"Only SELECT queries are allowed"

• You're trying to run a write query in read-only mode
• Add --allow-dangerous-queries true if write access is needed

"No database selected"

• The specified database doesn't exist
• The user doesn't have access to the database
• Check SHOW DATABASES; to see available databases

"Table doesn't exist"

• Verify you're querying the correct database
• Add the database parameter if using multiple databases
• Use SELECT DATABASE() to check current context

"Failed to acquire connection"

• Connection pool is exhausted
• Wait a moment and retry

Tool not appearing in Claude Desktop

1. Verify the path to the binary is absolute (not relative)
2. Check Claude Desktop logs for errors
3. Restart Claude Desktop completely (not just reload)
4. Ensure the server process starts without errors when run manually

Development

Project Structure

mcp-server-mysql/
├── src/
│   ├── main.rs          # Main server implementation
│   ├── config.rs        # Configuration handling
│   ├── db.rs            # Database operations
│   ├── rpc.rs           # RPC protocol handling
│   └── server.rs        # Server initialization
├── tests/               # Test files
├── Cargo.toml           # Rust dependencies
├── Cargo.lock          # Locked dependency versions
└── README.md           # This file

Building for Development

cargo build
cargo run -- --help

Running Tests

cargo test

Code Quality

# Format code
cargo fmt

# Run linter
cargo clippy

# Check for issues
cargo check

Development Conventions

The project follows standard Rust best practices:

• Code formatting: cargo fmt
• Linting: cargo clippy
• Testing: cargo test

Deployment Guide

Quick Deployment

1. Test Connection

./mcp-server-mysql \
  --username your_user \
  --password your_pass \
  --database your_db

Press Ctrl+C to exit after seeing "MCP MySQL Server started".

2. Configure Claude Desktop

Edit your Claude config file and add the server configuration (see Quick Start section).

3. Restart Claude Desktop

Close and reopen Claude Desktop completely.

Production Deployment Tips

Performance

• The binary is optimized with --release flag
• Connection pooling is configured (max 5 connections)
• Automatic retry logic for transient failures

Monitoring

Server logs go to stderr. Capture them with:

./mcp-server-mysql --username user --password pass --database db 2>> server.log

Log levels:

• INFO: Connection events, tool calls
• DEBUG: Detailed query information
• WARN: Non-fatal issues
• ERROR: Failures and errors

Systemd Service (Optional)

For long-running deployments, create /etc/systemd/system/mcp-mysql.service:

[Unit]
Description=MySQL MCP Server
After=network.target mysql.service

[Service]
Type=simple
User=mcp-user
ExecStart=/usr/local/bin/mcp-server-mysql --username mcp_user --password secret --database production
Restart=on-failure
RestartSec=5s
StandardOutput=journal
StandardError=journal

[Install]
WantedBy=multi-user.target

Enable and start:

sudo systemctl enable mcp-mysql
sudo systemctl start mcp-mysql
sudo systemctl status mcp-mysql

Upgrading

# Backup current version
cp /usr/local/bin/mcp-server-mysql /usr/local/bin/mcp-server-mysql.backup

# Replace with new version
cp mcp-server-mysql /usr/local/bin/

# Restart services
sudo systemctl restart mcp-mysql  # If using systemd
# Or restart Claude Desktop

Rollback

# Restore previous version
cp /usr/local/bin/mcp-server-mysql.backup /usr/local/bin/mcp-server-mysql

# Or checkout previous git tag
git checkout v0.1.0
cargo build --release

Contributing

Contributions are welcome! Please ensure:

• Code follows Rust best practices
• All tests pass
• Documentation is updated
• Commit messages are clear and descriptive

License

Apache-2.0

Support

For issues, questions, or contributions, please open an issue on the project repository.

Version: 0.2.0 | Release Date: 2025-01-XX | Protocol: MCP 2025-03-26 | Platform: Linux x86_64 | Status: Production Ready ✅