A lightweight service that enables AI assistants to execute AWS CLI commands (in safe containerized environment) through the Model Context Protocol (MCP). Bridges Claude, Cursor, and other MCP-aware AI tools with AWS CLI for enhanced cloud infrastructure management.
Configuration
View docs{
"mcpServers": {
"alexei-led-aws-mcp-server": {
"command": "uvx",
"args": [
"aws-mcp"
],
"env": {
"AWS_REGION": "us-west-2",
"AWS_PROFILE": "my-profile",
"AWS_CONFIG_FILE": "~/.aws/config",
"AWS_MCP_SANDBOX": "auto",
"AWS_MCP_TIMEOUT": "300",
"AWS_MCP_TRANSPORT": "stdio",
"AWS_MCP_MAX_OUTPUT": "100000",
"AWS_MCP_SANDBOX_CREDENTIALS": "both",
"AWS_SHARED_CREDENTIALS_FILE": "~/.aws/credentials"
}
}
}
}You can empower Claude to execute AWS CLI commands on your behalf by running an MCP server that wraps the AWS CLI itself. This setup lets Claude access a wide range of AWS services through simple tools, while you retain control with your IAM policies and credentials.
You interact with the AWS MCP Server through an MCP client. Claude can learn common AWS commands on demand using the built-in help tool, then execute those commands against your AWS account. The server acts as a bridge, running the AWS CLI commands locally or in a container, and returning results through the MCP client.
Prerequisites you need before starting:
- AWS CLI installed
- AWS credentials configured
- uvx installed for the MCP runtime (for the standard local setup)1. Create the MCP configuration to run the server locally using uvx. Use the following configuration snippet exactly as shown to connect to the MCP server from Claude.
{
"mcpServers": {
"aws": {
"command": "uvx",
"args": ["aws-mcp"]
}
}
}2. If you plan to use a specific AWS profile, extend the configuration to pass the profile via environment variables as shown.
{
"mcpServers": {
"aws": {
"command": "uvx",
"args": ["aws-mcp"],
"env": {
"AWS_PROFILE": "my-profile"
}
}
}
}Get documentation for any AWS CLI command.
Execute AWS CLI commands with optional pipes (e.g., jq, grep).