AWS Agent Toolkit

Overview

AWS Agent Toolkit is the official collection of Model Context Protocol (MCP) servers from Amazon Web Services, designed to enable AI agents to interact with AWS services programmatically. It provides a standardized interface for AI coding assistants to provision, manage, and interact with AWS infrastructure through natural language, bridging the gap between AI agents and cloud operations.

The toolkit includes multiple purpose-built MCP servers covering core AWS services, enabling agents to perform tasks ranging from EC2 instance management and S3 operations to Lambda function deployment and CloudWatch monitoring. As AWS's official entry into the MCP ecosystem, it represents a significant milestone in making cloud infrastructure accessible through AI agent interfaces.

Features

• Multi-service Support: MCP servers for EC2, S3, Lambda, DynamoDB, CloudWatch, IAM, and more.
• Official AWS Integration: Built and maintained by AWS, ensuring compatibility with the latest service APIs.
• Natural Language Infrastructure Management: Provision and manage AWS resources through conversational AI.
• Security-aware: Leverages AWS IAM for authentication and authorization, respecting existing permission boundaries.
• Extensible Architecture: Modular design allows adding support for additional AWS services as MCP servers.

Installation

Using npx

npx @aws/mcp-ec2
npx @aws/mcp-s3
npx @aws/mcp-lambda

Using pip

pip install aws-mcp-ec2 aws-mcp-s3 aws-mcp-lambda

Configuration

Add the AWS MCP servers to your Claude Desktop configuration:

{
  "mcpServers": {
    "aws-ec2": {
      "command": "npx",
      "args": ["-y", "@aws/mcp-ec2"],
      "env": {
        "AWS_REGION": "us-east-1",
        "AWS_PROFILE": "default"
      }
    },
    "aws-s3": {
      "command": "npx",
      "args": ["-y", "@aws/mcp-s3"],
      "env": {
        "AWS_REGION": "us-east-1",
        "AWS_PROFILE": "default"
      }
    },
    "aws-lambda": {
      "command": "npx",
      "args": ["-y", "@aws/mcp-lambda"],
      "env": {
        "AWS_REGION": "us-east-1",
        "AWS_PROFILE": "default"
      }
    }
  }
}

Available Tools

Server	Tool	Description
aws-ec2	list_instances	List EC2 instances with optional filters
aws-ec2	describe_instance	Get detailed information about a specific instance
aws-ec2	start_instance	Start a stopped EC2 instance
aws-ec2	stop_instance	Stop a running EC2 instance
aws-s3	list_buckets	List all S3 buckets in the account
aws-s3	list_objects	List objects in an S3 bucket
aws-s3	get_object	Retrieve an object from S3
aws-s3	put_object	Upload an object to S3
aws-lambda	list_functions	List Lambda functions
aws-lambda	invoke_function	Invoke a Lambda function
aws-lambda	update_function_code	Update a Lambda function's code

Usage Examples

Via Claude Desktop

User: "List all my EC2 instances in us-east-1"
Agent: [Calls list_instances with region filter]
Agent: "You have 3 EC2 instances in us-east-1: web-server-01 (running), db-server-01 (stopped), and staging-01 (running)."

User: "Deploy the latest build to S3"
Agent: [Calls s3 upload with build artifacts]
Agent: "Deployed successfully to my-app-bucket/releases/v2.1.0/"

Claude Desktop Setup

1. Install the AWS CLI and configure your credentials:

   aws configure
   

2. Add the desired MCP servers to your claude_desktop_config.json.

3. Restart Claude Desktop and verify the tools appear in the MCP tool list.

Pros

• ✅ Official AWS support with guaranteed API compatibility
• ✅ Secure with existing IAM roles and policies
• ✅ Covers core AWS services comprehensively
• ✅ Easy to extend with new service modules
• ✅ Active development from AWS

Cons

• ❌ Requires AWS credentials and CLI configuration
• ❌ Limited to AWS ecosystem only
• ❌ Newer project with evolving documentation
• ❌ Each service requires a separate MCP server installation

When to Use

Use AWS Agent Toolkit when:

• You regularly manage AWS infrastructure and want AI assistance
• You're building agents that need to provision or interact with cloud resources
• You want natural language interfaces for DevOps workflows
• You're already invested in the AWS ecosystem

Resources

• GitHub Repository
• AWS Documentation
• AWS MCP Announcement