AWS Billing

You can run an MCP server that provides AWS billing and cost analysis tools inside Claude Desktop. It offers cost analysis, usage comparison, trend analysis, anomaly detection, and cost ranking, with flexible configuration to test with mock data or real AWS data.

How to use

Connect your MCP client to the AWS Billing MCP Server through the standard MCP protocol. You will interact with the server by issuing tool requests such as analyzing costs, comparing usage across periods, identifying cost trends, detecting anomalies, and ranking cost drivers. Use natural language prompts like “Show me EC2 costs for last month” or “Compare this month to last month” to trigger the appropriate analysis. The server handles data retrieval, analysis, and presentation of results within Claude Desktop.

How to install

Prerequisites you need before installing include: a currently supported version of Node.js and npm, a working Claude Desktop setup, and optional AWS credentials if you want to access real AWS data.

1. Install dependencies and build the MCP server.

npm install
npm run build

Configure and run the MCP server in Claude Desktop

Prepare the Claude Desktop configuration so it can start the AWS Billing MCP Server. The configuration below runs locally with environment variables for AWS access. This setup uses mock data by default if credentials are not provided.

{
  "mcpServers": {
    "aws_billing_mcp": {
      "command": "node",
      "args": ["/path/to/your/project/start-mcp-server.js"],
      "env": {
        "AUTH_ENABLED": "false",
        "LOG_LEVEL": "info",
        "AWS_ACCESS_KEY_ID": "your-access-key-id",
        "AWS_SECRET_ACCESS_KEY": "your-secret-access-key",
        "AWS_REGION": "us-east-1"
      }
    }
  }
}

Start and test with Claude

Place the above configuration in Claude Desktop under your profile, then restart Claude Desktop to load the MCP server. If you want to verify the setup, ask Claude to analyze AWS costs for the last month.

Notes on testing

• Run unit tests with the project’s test script if you have one configured in your environment.

• Check Claude Desktop MCP server logs for startup messages and any errors during initialization.

Configuration

Use environment variables to control AWS access and feature flags. The example below demonstrates the explicit configuration you can place in your Claude Desktop setup.

{
  "mcpServers": {
    "aws_billing_mcp": {
      "command": "node",
      "args": ["/path/to/your/project/start-mcp-server.js"],
      "env": {
        "AUTH_ENABLED": "false",
        "LOG_LEVEL": "info",
        "AWS_ACCESS_KEY_ID": "your-access-key-id",
        "AWS_SECRET_ACCESS_KEY": "your-secret-access-key",
        "AWS_REGION": "us-east-1"
      }
    }
  }
}

Security

Authentication is disabled by default for local development. AWS credentials are stored in the configuration and are not saved to a database by the MCP server. For production deployments, you can enable Google SSO as an additional security layer.

Testing and health

• Access the health endpoint at http://localhost:37291/health to verify the MCP server is running.

• Review console logs for startup messages and potential errors.

Files and structure

Key files and folders you will interact with include the source code under src/, the compiled output under dist/, and the configuration template you place in Claude Desktop.

Available tools

analyze_costs

Analyze AWS costs with filtering to break down expenses by service, region, tag, or time period.

compare_usage

Compare usage and costs between two or more time periods to spot differences.

analyze_trends

Identify cost trends over time to forecast future spending.

detect_anomalies

Detect unusual cost patterns that may indicate misconfigurations or anomalies.

rank_costs

Identify top cost drivers by service, region, or other dimensions.