A Model Context Protocol (MCP) server that provides tools to interact with AWS services through Claude Desktop.

• S3: List buckets, list objects
• EC2: Describe instances, security groups, VPCs
• RDS: Describe database instances
• Cost Explorer: Get cost and usage reports
• CloudWatch: Retrieve metric statistics
• Generic AWS SDK: Access any AWS operation via aws_sdk_wrapper
• Vector Store: Optional document ingestion and search capabilities

• Python 3.12+
• AWS credentials configured (~/.aws/credentials)
• Claude Desktop installed

1. Install the package:

   pip install aws-mcp-server

2. Configure Claude Desktop: Add to your claude_desktop_config.json:

   {
       "mcpServers": {
           "aws-mcp-server": {
               "command": "aws-mcp-server"
           }
       }
   }

3. Configure AWS credentials:

   aws configure
   # OR manually edit ~/.aws/credentials:

   [default]
   aws_access_key_id = YOUR_ACCESS_KEY
   aws_secret_access_key = YOUR_SECRET_KEY
   region = us-east-1

# Clone and install
git clone <repository-url>
cd aws-mcp-server
uv sync

# Run locally
uvx .

Set environment variables for customization:

export AWS_MCP_PORT=8888
export AWS_MCP_DEBUG=true
export ENABLE_VECTOR_STORE=true

• Deployment Guide - Production deployment with AWS App Runner
• Vector Store Guide - Document ingestion and search setup
• Workflow Guide - Development workflows and patterns

MIT License - see LICENSE file for details.