Skip to main content
Glama

claude-aws-mcp

PyPI version Python License: MIT

An MCP server that lets you control AWS resources directly from Claude Desktop or Claude Code using natural language.


Installation

pip install claude-aws-mcp

Setup

Step 1. Create a .env file in your working directory:

AWS_ACCESS_KEY_ID=your_access_key
AWS_SECRET_ACCESS_KEY=your_secret_key
AWS_DEFAULT_REGION=ap-northeast-2

Step 2. Run the MCP server:

aws-mcp

Claude Desktop Integration

Config file location:

OS

Path

Windows

%APPDATA%\Claude\claude_desktop_config.json

Mac

~/Library/Application Support/Claude/claude_desktop_config.json

Add this to claude_desktop_config.json:

{
  "mcpServers": {
    "aws": {
      "command": "aws-mcp",
      "env": {
        "AWS_ACCESS_KEY_ID": "your_access_key",
        "AWS_SECRET_ACCESS_KEY": "your_secret_key",
        "AWS_DEFAULT_REGION": "ap-northeast-2"
      }
    }
  }
}

Restart Claude Desktop — the AWS tools will appear automatically.


Claude Code Integration

claude mcp add aws aws-mcp

Or add directly to .claude/settings.json:

{
  "mcpServers": {
    "aws": {
      "command": "aws-mcp"
    }
  }
}

Available Tools

Security Groups

Tool

Description

list_security_groups

List all security groups with inbound/outbound rules

get_my_public_ip

Get the current public IP address

add_my_ip_to_security_group

Add current IP to a specified security group

remove_ip_from_security_group

Remove a specific CIDR rule (supports port ranges e.g. 0-65535)

create_security_group

Create a new security group

EC2 Instances

Tool

Description

list_ec2_instances

List instances with state, IP, and type

get_instance_details

Get detailed info for a specific instance

start_ec2_instance

Start an instance

stop_ec2_instance

Stop an instance

reboot_ec2_instance

Reboot an instance

create_ec2_instance

Create a new instance

terminate_ec2_instance

Permanently delete an instance

get_instance_console_output

Get boot log (console output)

Elastic IP

Tool

Description

list_elastic_ips

List allocated EIPs with associated instance info

allocate_elastic_ip

Allocate a new EIP

associate_elastic_ip

Associate an EIP with an instance

disassociate_elastic_ip

Disassociate an EIP

release_elastic_ip

Release an EIP

Utilities

Tool

Description

get_ssh_command

Generate SSH command (auto-detects username from AMI)

add_instance_tag

Add or update instance tags

change_instance_type

Change instance type (auto stops if running)

S3

Tool

Description

list_s3_buckets

List all buckets with region

list_s3_objects

List objects in a bucket (supports prefix filter)

create_s3_bucket

Create a new bucket

delete_s3_object

Delete a file from a bucket

get_s3_presigned_url

Generate a temporary download URL

Bedrock AI

Tool

Description

list_bedrock_models

List available foundation models (supports provider filter)

invoke_bedrock_claude

Invoke Claude model (default: Claude 3.5 Sonnet v2)

invoke_bedrock_model_raw

Invoke any Bedrock model with raw JSON

list_bedrock_knowledge_bases

List Knowledge Bases

query_bedrock_knowledge_base

RAG query against a Knowledge Base

Infrastructure Info

Tool

Description

list_key_pairs

List available key pairs

list_vpcs

List VPCs and subnets

list_available_amis

List latest official AMIs (Amazon Linux / Ubuntu / Windows)


Example Prompts

Once connected to Claude, just type naturally:

Show me all security groups
Add my current IP to sg-0abc1234 on port 22
Remove 0.0.0.0/0 from sg-0abc1234 on port range 0-65535
List all running EC2 instances
Stop instance i-0abc1234
List all S3 buckets
Generate a presigned URL for my-bucket/report.pdf

IAM Permissions Required

Service

Recommended Policy

EC2 / Security Groups

AmazonEC2FullAccess

S3

AmazonS3FullAccess

Bedrock

AmazonBedrockFullAccess


Troubleshooting

aws-mcp command not found

pip install --upgrade claude-aws-mcp

AWS authentication errors

  • Check that your .env values are correct

  • Confirm the IAM user has the required permissions


Latest Blog Posts

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/chyang222/MCP_Project_AWS'

If you have feedback or need assistance with the MCP directory API, please join our Discord server