home / mcp / aws mcp server

AWS MCP Server

Provides interactive AWS access via natural language with multi-profile and multi-region support.

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "arunsanna-aws-sage": {
      "command": "python",
      "args": [
        "/path/to/your/aws-mcp/run_aws_mcp.py"
      ]
    }
  }
}

You can run a Python-based MCP server that lets Claude interact with your AWS environment through natural language queries and commands. It supports multiple AWS profiles, multi-region work, and local execution with your AWS credentials, making it easy to manage AWS resources during conversations.

How to use

Start by installing and running the MCP server locally, then connect it to your Claude client to enable natural language AWS operations. You can operate directly from the command line or configure Claude Desktop to launch the server as a background process. The server understands natural language requests and maps them to appropriate AWS actions via boto3, returning readable results.

How to install

Prerequisites you need to prepare before installation:

Python
Claude Desktop
AWS credentials configured locally (the ~\/\.aws directory)

1. Install the MCP server in editable mode from your project directory.

pip install -e .

Additional sections

Configuration and runtime details are provided to allow smooth integration with Claude Desktop and your AWS setup. The server supports using the AWS credentials you have locally and can work with multiple AWS profiles across different regions.

CLI usage patterns give you quick access to listing credentials, selecting a profile, and performing AWS operations. You can also run operations by providing a natural language command or a code-like request.

# Show available commands
python -m aws_mcp --help

# List available AWS profiles
python -m aws_mcp list-credentials

# Select an AWS profile
python -m aws_mcp select-profile myprofile

# Run AWS operations
python -m aws_mcp run-aws-code "list all S3 buckets"

# Use pretty formatting for more readable output
python -m aws_mcp --format pretty list-credentials

Configuration and integration with Claude Desktop

To connect Claude Desktop to the MCP server, configure Claude’s developer settings to launch the MCP server as a stdio-based process.

{
  "mcpServers": {
    "aws-mcp": {
      "command": "/path/to/your/python",
      "args": ["/path/to/your/aws-mcp/run_aws_mcp.py"]
    }
  }
}

Examples of Claude interactions

Start by selecting an AWS profile, then ask for actions such as listing resources, describing services, or running AWS operations across regions.

Troubleshooting

If you run into issues, check the log files for clues about startup problems or runtime errors. Typical locations include user-specific log paths that Claude uses for MCP servers.

Notes and tips

- The AWS MCP supports dynamic execution of AWS operations via boto3, parsing natural language or code-like requests to determine the AWS service and operation, then formatting results for readability.

Available tools

nlp_parse_service

Parses natural language requests to identify the AWS service and operation to perform.

execute_boto3

Executes the corresponding boto3 command for the chosen AWS service and operation.

format_results

Formats results into a readable table or structured output.

error_handling

Handles errors gracefully and provides actionable feedback.