PostgreSQL Operations and Database Monitoring

The PostgreSQL Operations and Database Monitoring MCP Server lets you interact with your PostgreSQL databases using natural language queries.

Instead of writing complex SQL commands, you ask questions like “Show me slow queries” or “Analyze table bloat” and get immediate, actionable results.

This MCP server supports PostgreSQL versions 12 through 17 and works right out of the box with automatic version detection. It’s production-safe with read-only operations and works with RDS and Aurora databases using regular user permissions.

Key Features

• 🚀 Zero configuration setup with automatic PostgreSQL version detection
• 💬 Natural language queries instead of complex SQL commands
• 🔒 Read-only operations that work with production databases
• 📊 Comprehensive monitoring including performance analysis and bloat detection
• 🔍 Schema exploration with detailed relationship mapping
• ⚡ Real-time connection and lock monitoring
• 📈 Version-aware statistics with enhanced features on PostgreSQL 16+
• 🔧 Multi-database operations across your entire PostgreSQL environment

Practical Use Cases

• Performance troubleshooting during incidents when you need quick insights without writing complex queries
• Routine database health checks that would normally require multiple monitoring tools
• Schema discovery and relationship mapping for unfamiliar databases
• Capacity planning and maintenance scheduling through bloat analysis and vacuum monitoring

Implementation Guide

1. Clone the repository and check the environment configuration:

git clone https://github.com/call518/MCP-PostgreSQL-Ops.git
cd MCP-PostgreSQL-Ops
cp .env.example .env

2. The default configuration uses an included test database, but you can modify the .env file to point to your own PostgreSQL instance:

# Commons
PYTHONPATH=/app/src
# MCP Server Configuration
FASTMCP_TYPE=streamable-http
FASTMCP_HOST=0.0.0.0
FASTMCP_PORT=8080
MCP_LOG_LEVEL=INFO
### PostgreSQL Configuration
PGSQL_VERSION=16
PGDATA=/data/db ### Don't Edit
POSTGRES_MAX_CONNECTIONS=200
#### PostgreSQL Host Configuration:
#--------------------------------------------
POSTGRES_HOST=host.docker.internal
POSTGRES_PORT=15432
POSTGRES_USER=postgres
POSTGRES_PASSWORD=changeme!@34
POSTGRES_DB=ecommerce
#--------------------------------------------
# Docker Env.
DOCKER_EXTERNAL_PORT_OPENWEBUI=3003
DOCKER_EXTERNAL_PORT_MCP_SERVER=18003
DOCKER_EXTERNAL_PORT_MCPO_PROXY=8003

3. Start the containers:

docker-compose up -d

4. Access OpenWebUI at http://localhost:3003 and register the tool using the address http://localhost:8003/postgresql-ops.

5. Connect your preferred AI model (Ollama or OpenAI), and you’re ready to start querying.

6. For Claude Desktop integration, add this to your configuration:

{
  "mcpServers": {
    "postgresql-ops": {
      "command": "uvx",
      "args": ["--python", "3.11", "mcp-postgresql-ops"],
      "env": {
        "POSTGRES_HOST": "127.0.0.1",
        "POSTGRES_PORT": "5432",
        "POSTGRES_USER": "your_user",
        "POSTGRES_PASSWORD": "your_password",
        "POSTGRES_DB": "your_database"
      }
    }
  }
}

FAQs

Q: Does this work with managed PostgreSQL services like RDS or Aurora?
A: Yes, it works perfectly with managed services. The server uses read-only operations that work with regular user permissions. For advanced query analysis, you’ll need to enable pg_stat_statements through your DB parameter group.

Q: What PostgreSQL configuration is required for full functionality?
A: Most tools work without any configuration changes. For function statistics, set track_functions = ‘pl’ in your postgresql.conf. For accurate I/O timing, enable track_io_timing = on. These can be set dynamically on managed services using ALTER SYSTEM commands.

Q: How does it handle different PostgreSQL versions?
A: The server automatically detects your PostgreSQL version and adjusts its queries accordingly. Features that require specific versions (like enhanced I/O stats in PG16+) gracefully fall back to compatible alternatives on older versions.

Q: Is this safe for production databases?
A: Absolutely. All operations are read-only and the server only requires standard monitoring permissions. No data modification operations are possible through the MCP interface.

Q: What if I don’t have the required extensions installed?
A: The majority of tools work without any extensions. Only the advanced query analysis features require pg_stat_statements or pg_stat_monitor. You’ll still get comprehensive monitoring capabilities even without these extensions.