RDS MCP Server

A Model Context Protocol (MCP) server for monitoring and analyzing Amazon RDS(MySQL, PostgreSQL) instances information, metrics and slowquery logs.

Installation

Install directly from GitHub using pipx:

# Install
pipx install git+https://github.com/neoai-agent/rds-mcp.git

# Or run without installation
pipx run git+https://github.com/neoai-agent/rds-mcp.git

Quick Start

Authentication Options

The server supports multiple AWS authentication methods:

Option 1: IAM Roles (Recommended for EC2/ECS)

When running on AWS infrastructure with IAM roles attached, you can omit AWS credentials:

rds-mcp --openai-api-key "YOUR_OPENAI_API_KEY" --region "YOUR_AWS_REGION"

Option 2: AWS Access Keys

For local development or when IAM roles are not available:

rds-mcp --access-key "YOUR_AWS_ACCESS_KEY" --secret-access-key "YOUR_AWS_SECRET_KEY" --region "YOUR_AWS_REGION" --openai-api-key "YOUR_OPENAI_API_KEY"

Option 3: Environment Variables

You can also set AWS credentials via environment variables:

rds-mcp --access-key "YOUR_AWS_ACCESS_KEY" --secret-access-key "YOUR_AWS_SECRET_KEY" --region "YOUR_AWS_REGION" --openai_api_key "YOUR_OPENAI_API_KEY"

Note: When using IAM roles, the server will automatically use the default AWS credential chain, which includes IAM roles, environment variables, and AWS credentials file.

Available Tools

The server provides the following tools for RDS instance management and monitoring:

1. Get RDS instance details:

await get_db_info(
    database_name="your-db-instance",
    region="your-aws-region"
)

2. Get database metrics:

await get_database_metrics(
    database_name="your-db-instance",
    time_range_minutes=30
)

3. Get slow queries:

await get_database_queries(
    database_name="your-db-instance",
    time_range_minutes=30
)

4. Get instance performance metrics:

await get_top_rds_load(
    database_name="your-db-instance",
    time_range_minutes=30
)

Development

For development setup:

git clone https://github.com/neoai-agent/rds-mcp.git
cd rds-mcp
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate
pip install -e ".[dev]"

License

MIT License - See LICENSE file for details