aws

This AWS MCP server exposes natural-language capable commands to interact with AWS resources through the Model Context Protocol. Once running, you can query and manage AWS resources without switching to the AWS Console or CLI, by simply asking in plain language (e.g., list profiles, show EC2 instances, inspect S3 buckets, or check Lambda deployments). The server operates locally with your AWS credentials and supports multiple AWS profiles and multi-region setups, enabling contextual actions across accounts and regions. Start Claude Desktop (or your MCP host) and point it to the AWS MCP entry so your LLM can send MCP requests to the server and receive structured responses.

Available capabilities include querying and modifying AWS resources via natural language, supporting multiple AWS profiles and SSO, multi-region access, secure local credential handling, and local execution using your existing AWS credentials. Typical use cases involve inventorying resources, filtering by regions or tags, and performing read-only queries or scripted actions through LLM prompts that translate to AWS API calls.

Prerequisites

• Node.js 18+ installed on your machine
• Claude Desktop application (or an MCP host) to connect to the MCP server
• Locally configured AWS credentials (AWS Configure) in ~/.aws/

1. Clone the repository

git clone https://github.com/ihatesea69/AWS-MCP
cd aws-mcp

2. Install dependencies You can use pnpm or npm:

pnpm install
# or
npm install

3. Start the MCP server Assuming you want to run via npm with the provided configuration:

# This uses the npm-based start script in the project root

4. Configure Claude Desktop (or your MCP client) Open Claude Desktop and navigate to Settings -> Developer -> Edit Config, and add the following entry to claude_desktop_config.json under mcpServers:

{
  "mcpServers": {
    "aws": {
      "command": "npm", // or pnpm
      "args": [
        "--silent",
        "--prefix",
        "/Users/<YOUR USERNAME>/aws-mcp",
        "start"
      ]
    }
  }
}

Replace /Users/<YOUR USERNAME>/aws-mcp with the actual path to your project directory.

5. Restart Claude Desktop and test Restart Claude Desktop. If installed correctly, you should see a successful connection message with MCP. Then try natural language prompts like: "List available AWS profiles" or "Show me EC2 instances in us-east-1".

Tips and caveats:

• Ensure your AWS credentials are configured locally (in ~/.aws) and accessible to the MCP server. The server uses these credentials to interact with AWS APIs.
• If you use nvm, you can configure an alternative start command that points directly to node and the project’s TSX/TS code, as shown in the README. This allows running with a specific Node.js version.
• For Claude Desktop users, the provided config expects the server to be started from the project root using npm start (or pnpm start). Adjust the --prefix path to match your environment.
• If you encounter connection issues, verify that the MCP server process is running and that Claude Desktop is correctly configured to reach the local server endpoint.
• You can run multiple MCP servers by adding more entries under mcpServers with unique server names (e.g., aws, gcp, azure).