aws mcp server

A lightweight service that enables AI assistants to execute AWS CLI commands (in safe containerized environment) through the Model Context Protocol (MCP). Bridges Claude, Cursor, and other MCP-aware AI tools with AWS CLI for enhanced cloud infrastructure management.

alexei-led

Python

AWS Model Context Protocol (MCP) Server

A lightweight service that enables AI assistants to execute AWS CLI commands through the Model Context Protocol (MCP).

Overview

The AWS MCP Server provides a bridge between MCP-aware AI assistants (like Claude Desktop, Cursor, Windsurf) and the AWS CLI. It enables these assistants to:

Retrieve AWS CLI documentation (aws_cli_help) - Get detailed help on AWS services and commands
Execute AWS CLI commands (aws_cli_pipeline) - Run commands with Unix pipes and receive formatted results optimized for AI consumption

flowchart LR
    AI[AI Assistant] <-->|MCP Protocol| Server[AWS MCP Server]
    Server <-->|Subprocess| AWS[AWS CLI]
    AWS <-->|API| Cloud[AWS Cloud]

Demo

The video demonstrates using Claude Desktop with AWS MCP Server to create a new AWS EC2 instance with AWS SSM agent installed.

Features

Command Documentation - Detailed help information for AWS CLI commands
Command Execution - Execute AWS CLI commands and return human-readable results
Unix Pipe Support - Filter and transform AWS CLI output using standard Unix pipes and utilities
AWS Resources Context - Access to AWS profiles, regions, account information, and environment details via MCP Resources
Prompt Templates - Pre-defined prompt templates for common AWS tasks following best practices
Docker Integration - Simple deployment through containerization with multi-architecture support (AMD64/x86_64 and ARM64)
AWS Authentication - Leverages existing AWS credentials on the host machine

Requirements

Docker (default) or Python 3.13+ (and AWS CLI installed locally)
AWS credentials configured

Getting Started

Note: For security and reliability, running the server inside a Docker container is the strongly recommended method. Please review the Security Considerations section for important considerations.

Run Server Option 1: Using Docker (Recommended)

# Clone repository
git clone https://github.com/alexei-led/aws-mcp-server.git
cd aws-mcp-server

# Build and run Docker container
docker compose -f deploy/docker/docker-compose.yml up -d

The Docker image supports both AMD64/x86_64 (Intel/AMD) and ARM64 (Apple Silicon M1-M4, AWS Graviton) architectures.

Note: The official image from GitHub Packages is multi-architecture and will automatically use the appropriate version for your system.
# Use the latest stable version
docker pull ghcr.io/alexei-led/aws-mcp-server:latest

# Or pin to a specific version (recommended for production)
docker pull ghcr.io/alexei-led/aws-mcp-server:1.0.0
Docker Image Tags:

latest: Latest stable release

x.y.z (e.g., 1.0.0): Specific version

sha-<commit-sha>: Development builds, tagged with Git commit SHA (e.g., sha-gb697684)

Run Server Option 2: Using Python

Use with Caution: Running natively requires careful environment setup and carries higher security risks compared to the recommended Docker deployment. Ensure you understand the implications outlined in the Security Considerations section.

# Clone repository
git clone https://github.com/alexei-led/aws-mcp-server.git
cd aws-mcp-server

# Set up virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install in development mode
pip install -e .

# Run the server
python -m aws_mcp_server

Configuration

The AWS MCP Server can be configured using environment variables:

Environment Variable	Description	Default
`AWS_MCP_TIMEOUT`	Command execution timeout in seconds	300
`AWS_MCP_MAX_OUTPUT`	Maximum output size in characters	100000
`AWS_MCP_TRANSPORT`	Transport protocol to use (“stdio” or “sse”)	stdio
`AWS_PROFILE`	AWS profile to use	default
`AWS_REGION`	AWS region to use	us-east-1
`AWS_MCP_SECURITY_MODE`	Security mode (“strict” or “permissive”)	strict
`AWS_MCP_SECURITY_CONFIG`	Path to custom security configuration file	“”

Important: Securely manage the AWS credentials provided to the server, whether via mounted ~/.aws files or environment variables. Ensure the credentials follow the principle of least privilege as detailed in the Security Considerations section. When running via Docker, ensure these variables are passed correctly to the container environment (e.g., using docker run -e VAR=value ...).

Security Considerations

Security is paramount when executing commands against your AWS environment. While AWS MCP Server provides functionality, you are responsible for configuring and running it securely. Please adhere strictly to the following:

1. Recommended Deployment: Docker Container

Isolation: Running the server inside a Docker container is the strongly recommended and default deployment method. Containerization provides crucial filesystem and process isolation. Potentially destructive Unix commands (like rm, mv) executed via pipes, even if misused, will be contained within the ephemeral Docker environment and will not affect your host machine’s filesystem. The container can be easily stopped and recreated.
Controlled Environment: Docker ensures a consistent environment with necessary dependencies, reducing unexpected behavior.

2. AWS Credentials and IAM Least Privilege (Critical)

User Responsibility: You provide the AWS credentials to the server (via mounted ~/.aws or environment variables).
Least Privilege is Essential: The server executes AWS CLI commands using the credentials you provide. It is absolutely critical that these credentials belong to an IAM principal (User or Role) configured with the minimum necessary permissions (least privilege) for only the AWS actions you intend to perform through this tool.
- Do Not Use Root Credentials: Never use AWS account root user credentials.
- Regularly Review Permissions: Periodically audit the IAM permissions associated with the credentials.
Impact Limitation: Properly configured IAM permissions are the primary mechanism for limiting the potential impact of any command executed via the server, whether intended or unintended. Even if a command were manipulated, it could only perform actions allowed by the specific IAM policy.

3. Trusted User Model

The server assumes the end-user interacting with the MCP client (e.g., Claude Desktop, Cursor) is the same trusted individual who configured the server and provided the least-privilege AWS credentials. Do not expose the server or connected client to untrusted users.

4. Understanding Execution Risks (Current Implementation)

Command Execution: The current implementation uses shell features (shell=True in subprocess calls) to execute AWS commands and handle Unix pipes. While convenient, this approach carries inherent risks if the input command string were manipulated (command injection).
Mitigation via Operational Controls: In the context of the trusted user model and Docker deployment, these risks are mitigated operationally:
- The trusted user is assumed not to provide intentionally malicious commands against their own environment.
- Docker contains filesystem side-effects.
- Crucially, IAM least privilege limits the scope of any AWS action that could be executed.
Credential Exfiltration Risk: Despite containerization and IAM, a sophisticated command injection could potentially attempt to read the mounted credentials (~/.aws) or environment variables within the container and exfiltrate them (e.g., via curl). Strict IAM policies remain the most vital defense to limit the value of potentially exfiltrated credentials.

5. Network Exposure (SSE Transport)

If using the sse transport (which implies a network listener), ensure you bind the server only to trusted network interfaces (e.g., localhost) or implement appropriate network security controls (firewalls, authentication proxies) if exposing it more broadly. The default stdio transport does not open network ports.

6. Shared Responsibility Summary

AWS MCP Server provides the tool.
You, the user, are responsible for:
- Running it within the recommended secure Docker environment.
- Providing and securely managing least-privilege AWS credentials.
- Ensuring only trusted users interact with the server/client.
- Securing the network environment if applicable.

By strictly adhering to Docker deployment and meticulous IAM least-privilege configuration, you establish the necessary operational controls for using the AWS MCP Server securely with its current implementation.

Integrating with Claude Desktop

Configuration

To manually integrate AWS MCP Server with Claude Desktop:

Locate the Claude Desktop configuration file:
- macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
- Windows: %APPDATA%\Claude\claude_desktop_config.json

Edit the configuration file to include the AWS MCP Server:

{
  "mcpServers": {
    "aws-mcp-server": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "-v",
        "/Users/YOUR_USER_NAME/.aws:/home/appuser/.aws:ro",
        "ghcr.io/alexei-led/aws-mcp-server:latest"
      ]
    }
  }
}

Restart Claude Desktop to apply the changes
- After restarting, you should see a hammer 🔨 icon in the bottom right corner of the input box
- This indicates that the AWS MCP Server is available for use

flowchart TD
    subgraph "User Device"
        config[Edit claude_desktop_config.json]
        claude[Claude Desktop]
        docker[Docker Container]
        aws_creds[AWS Credentials]
    end
    
    subgraph "AWS Cloud"
        services[AWS Services]
    end
    
    subgraph "MCP Server Components"
        tools[AWS CLI Tools]
        resources[AWS Resources]
        templates[Prompt Templates]
    end
    
    config -->|Add MCP Server Config| claude
    claude -->|Docker Run Command| docker
    docker --- tools
    docker --- resources
    docker --- templates
    aws_creds -->|Mount Read-only| docker
    resources -.->|Read| aws_creds
    tools -->|API Calls| services

Example Interactions

Getting AWS CLI Documentation:

User: What are the available options for the AWS S3 cp command?

Claude: Let me get that information for you.

The AWS S3 cp command has the following options:

NAME
    cp - Copies a local file or S3 object to another location locally or in S3.

SYNOPSIS
    cp <LocalPath> <S3Uri> or <S3Uri> <LocalPath> or <S3Uri> <S3Uri> [--acl <value>] [--cache-control <value>] ...

Executing AWS CLI Commands:

User: List my S3 buckets

Claude: I'll list your S3 buckets for you.

2023-10-15 14:30:45 my-bucket-1
2023-11-20 09:15:32 my-bucket-2
2024-01-05 11:22:18 my-backup-bucket

Using Command Pipes:

User: List my S3 buckets that have "backup" in their name and show them in alphabetical order

Claude: Let me find that for you.

2024-01-05 11:22:18 my-backup-bucket
2024-03-01 09:44:12 weekly-backup-bucket
2024-03-15 13:10:57 database-backup-bucket

Accessing AWS Resources:

User: What AWS regions are available for me to use?

Claude: Let me check what AWS regions are available for you.

Available regions:
- us-east-1 (US East, N. Virginia) - Currently selected
- us-east-2 (US East, Ohio)
- us-west-1 (US West, N. California)
- us-west-2 (US West, Oregon)
- eu-west-1 (EU West, Ireland)
- eu-central-1 (EU Central, Frankfurt)
- ap-northeast-1 (Asia Pacific, Tokyo)
- ap-southeast-1 (Asia Pacific, Singapore)
- ap-southeast-2 (Asia Pacific, Sydney)

You're currently using the us-east-1 region.

Using Prompt Templates:

User: I need to perform a security audit of my S3 buckets

Claude: I'll help you audit the security of your S3 buckets.

Here are AWS CLI commands to perform a comprehensive security audit:

# List all buckets with public access settings
aws s3api list-buckets --query 'Buckets[*].Name' --output text | xargs -I {} aws s3api get-public-access-block --bucket {}

# Check bucket policies for public access
aws s3api list-buckets --query 'Buckets[*].Name' --output text | xargs -I {} aws s3api get-bucket-policy --bucket {} 2>/dev/null

# Identify buckets without encryption
aws s3api list-buckets --query 'Buckets[*].Name' --output text | xargs -I {} aws s3api get-bucket-encryption --bucket {} 2>/dev/null

# Check for buckets without logging enabled
aws s3api list-buckets --query 'Buckets[*].Name' --output text | xargs -I {} aws s3api get-bucket-logging --bucket {} 2>/dev/null

# Verify bucket versioning status
aws s3api list-buckets --query 'Buckets[*].Name' --output text | xargs -I {} aws s3api get-bucket-versioning --bucket {}

Available Prompt Templates

The AWS MCP Server includes the following pre-defined prompt templates:

Core Operations

Prompt	Description	Parameters
`create_resource`	Generate commands to create AWS resources with best practices	`resource_type`, `resource_name`
`resource_inventory`	Create comprehensive inventory of resources	`service`, `region` (optional)
`troubleshoot_service`	Generate commands to troubleshoot service issues	`service`, `resource_id`
`resource_cleanup`	Identify and safely clean up resources	`service`, `criteria` (optional)

Security & Compliance

Prompt	Description	Parameters
`security_audit`	Audit security settings for a specific AWS service	`service`
`security_posture_assessment`	Comprehensive security assessment across your AWS environment	None
`iam_policy_generator`	Create least-privilege IAM policies	`service`, `actions`, `resource_pattern` (optional)
`compliance_check`	Check compliance with standards	`compliance_standard`, `service` (optional)

Cost & Performance

Prompt	Description	Parameters
`cost_optimization`	Find cost optimization opportunities for a service	`service`
`performance_tuning`	Optimize and tune performance of AWS resources	`service`, `resource_id`

Infrastructure & Architecture

Prompt	Description	Parameters
`serverless_deployment`	Deploy serverless applications with best practices	`application_name`, `runtime` (optional)
`container_orchestration`	Set up container environments (ECS/EKS)	`cluster_name`, `service_type` (optional)
`vpc_network_design`	Design and implement secure VPC networking	`vpc_name`, `cidr_block` (optional)
`infrastructure_automation`	Automate infrastructure management	`resource_type`, `automation_scope` (optional)
`multi_account_governance`	Implement secure multi-account strategies	`account_type` (optional)

Reliability & Monitoring

Prompt	Description	Parameters
`service_monitoring`	Set up comprehensive monitoring	`service`, `metric_type` (optional)
`disaster_recovery`	Implement enterprise-grade DR solutions	`service`, `recovery_point_objective` (optional)

Security

The AWS MCP Server implements a comprehensive multi-layered approach to command validation and security:

Command Validation System

The server validates all AWS CLI commands through a three-layer system:

Basic Command Structure:
- Verifies commands start with ‘aws’ prefix and contain a valid service
- Ensures proper command syntax
Security-Focused Command Filtering:
- Dangerous Commands: Blocks commands that could compromise security
- Safe Patterns: Explicitly allows read-only operations needed for normal use
- Regex Pattern Matching: Prevents complex security risks with pattern matching
Pipe Command Security:
- Validates Unix commands used in pipes
- Restricts commands to a safe allowlist
- Prevents filesystem manipulation and arbitrary command execution

Default Security Configuration

The default security configuration focuses on preventing the following attack vectors:

1. Identity and Access Management (IAM) Risks

Blocked Command	Security Risk
`aws iam create-user`	Creates potential backdoor accounts with persistent access
`aws iam create-access-key`	Creates long-term credentials that can be stolen or misused
`aws iam attach-*-policy`	Potential privilege escalation via policy attachments
`aws iam put-user-policy`	Inline policies can grant excessive permissions
`aws iam create-policy`	Creating new policies with potentially dangerous permissions
`aws iam create-login-profile`	Creates console passwords for existing users
`aws iam deactivate-mfa-device`	Disables multi-factor authentication, weakening security
`aws iam update-assume-role-policy`	Modifies trust relationships, enabling privilege escalation

2. Audit and Logging Tampering

Blocked Command	Security Risk
`aws cloudtrail delete-trail`	Removes audit trail of AWS activity
`aws cloudtrail stop-logging`	Stops collecting activity logs, creating blind spots
`aws cloudtrail update-trail`	Can redirect or modify logging configuration
`aws config delete-configuration-recorder`	Disables AWS Config recording of resource changes
`aws guardduty delete-detector`	Disables threat detection capabilities

3. Sensitive Data Access and Protection

Blocked Command	Security Risk
`aws secretsmanager put-secret-value`	Modifies sensitive credentials
`aws secretsmanager delete-secret`	Removes sensitive credentials
`aws kms schedule-key-deletion`	Schedules deletion of encryption keys, risking data loss
`aws kms disable-key`	Disables encryption keys, potentially exposing data
`aws s3api put-bucket-policy`	Can create public S3 buckets, exposing data
`aws s3api delete-bucket-policy`	Removes protective policies from buckets

4. Network Security Risks

Blocked Command	Security Risk
`aws ec2 authorize-security-group-ingress`	Opens inbound network access, potential exposure
`aws ec2 authorize-security-group-egress`	Opens outbound network access, potential data exfiltration
`aws ec2 modify-instance-attribute`	Can alter security properties of instances

Many read-only operations that match these patterns are explicitly allowed via safe patterns:

All get-, list-, and describe- commands
All help commands (--help, help)
Simulation and testing commands (e.g., aws iam simulate-custom-policy)

Configuration Options

Security Modes:
- strict (default): Enforces all security validations
- permissive: Logs warnings but allows execution (use with caution)
Custom Configuration:
- Override default security rules via YAML configuration file
- Configure service-specific dangerous commands
- Define custom safe patterns and regex rules
- Environment variable: AWS_MCP_SECURITY_CONFIG
Execution Controls:
- Timeouts prevent long-running commands (default: 300 seconds)
- Output size limits prevent memory issues
- Environment variables: AWS_MCP_TIMEOUT, AWS_MCP_MAX_OUTPUT

Custom Security Rules Example

You can create custom security rules by defining a YAML configuration file:

# Example custom security configuration
# Save to a file and set AWS_MCP_SECURITY_CONFIG environment variable

# Dangerous commands to block
dangerous_commands:
  iam:
    # Only block specific IAM operations for your environment
    - "aws iam create-user"
    - "aws iam attach-user-policy"
  
  # Custom service restrictions for your organization
  lambda:
    - "aws lambda delete-function"
    - "aws lambda remove-permission"
  
  # Prevent accidental DynamoDB table deletion
  dynamodb:
    - "aws dynamodb delete-table"

# Safe patterns to explicitly allow
safe_patterns:
  # Global safe patterns
  general:
    - "--help"
    - "--dry-run"
  
  # Allow read operations on IAM
  iam:
    - "aws iam get-"
    - "aws iam list-"
  
  # Allow specific Lambda operations
  lambda:
    - "aws lambda list-functions"
    - "aws lambda get-function"

# Complex regex rules for security validation
regex_rules:
  general:
    # Prevent use of root credentials
    - pattern: "aws .* --profile\\s+root"
      description: "Prevent use of root profile"
      error_message: "Using the root profile is not allowed for security reasons"
  
  iam:
    # Block creation of admin users
    - pattern: "aws iam create-user.*--user-name\\s+.*admin.*"
      description: "Prevent creation of admin users"
      error_message: "Creating users with 'admin' in the name is restricted"
    
    # Prevent wildcards in IAM policies
    - pattern: "aws iam create-policy.*\"Effect\":\\s*\"Allow\".*\"Action\":\\s*\"\\*\".*\"Resource\":\\s*\"\\*\""
      description: "Prevent wildcards in policies"
      error_message: "Creating policies with '*' wildcards for both Action and Resource is not allowed"
  
  s3:
    # Prevent public bucket policies
    - pattern: "aws s3api put-bucket-policy.*\"Effect\":\\s*\"Allow\".*\"Principal\":\\s*\"\\*\""
      description: "Prevent public bucket policies"
      error_message: "Creating bucket policies with public access is restricted"

Security Examples

The system follows IAM best practices, focusing on preventing escalation of privilege:

# This command would be blocked (creates user)
aws iam create-user --user-name new-user
> Error: This command (aws iam create-user) is restricted for security reasons.

# This command would be blocked (attaches admin policy)
aws iam attach-user-policy --user-name any-user --policy-arn arn:aws:iam::aws:policy/AdministratorAccess
> Error: Attaching Administrator policies is restricted for security reasons.

# This command would be blocked (opens SSH port globally)
aws ec2 authorize-security-group-ingress --group-id sg-12345 --protocol tcp --port 22 --cidr 0.0.0.0/0
> Error: Opening non-web ports to the entire internet (0.0.0.0/0) is restricted.

# These commands are allowed (read-only operations)
aws iam list-users
aws s3 ls
aws ec2 describe-instances

Security Best Practices

Always use the default strict security mode in production
Follow the deployment recommendations in Security Considerations
Run with least-privilege AWS credentials
For custom configurations, focus on your security requirements

Development

Setting Up the Development Environment

# Install only runtime dependencies using pip
pip install -e .

# Install all development dependencies using pip
pip install -e ".[dev]"

# Or use uv for faster dependency management
make uv-install       # Install runtime dependencies
make uv-dev-install   # Install development dependencies

Makefile Commands

The project includes a Makefile with various targets for common tasks:

# Test commands
make test             # Run tests excluding integration tests
make test-unit        # Run unit tests only (all tests except integration tests)
make test-integration # Run integration tests only (requires AWS credentials)
make test-all         # Run all tests including integration tests

# Test coverage commands
make test-coverage    # Run tests with coverage report (excluding integration tests)
make test-coverage-all # Run all tests with coverage report (including integration tests)

# Linting and formatting
make lint             # Run linters (ruff check and format --check)
make lint-fix         # Run linters and auto-fix issues where possible
make format           # Format code with ruff

For a complete list of available commands, run make help.

Code Coverage

The project includes configuration for Codecov to track code coverage metrics. The configuration is in the codecov.yml file, which:

Sets a target coverage threshold of 80%
Excludes test files, setup files, and documentation from coverage reports
Configures PR comments and status checks

Coverage reports are automatically generated during CI/CD runs and uploaded to Codecov.

Integration Testing

Integration tests verify AWS MCP Server works correctly with actual AWS resources. To run them:

Set up AWS resources:
- Create an S3 bucket for testing
- Set the environment variable: export AWS_TEST_BUCKET=your-test-bucket-name
- Ensure your AWS credentials are configured

Run integration tests:

# Run all tests including integration tests
make test-all

# Run only integration tests
make test-integration

Or you can run the pytest commands directly:

# Run all tests including integration tests
pytest --run-integration

# Run only integration tests
pytest --run-integration -m integration

Troubleshooting

Authentication Issues: Ensure your AWS credentials are properly configured
Connection Errors: Verify the server is running and AI assistant connection settings are correct
Permission Errors: Check that your AWS credentials have the necessary permissions
Timeout Errors: For long-running commands, increase the AWS_MCP_TIMEOUT environment variable

Why Deploy with Docker

Deploying AWS MCP Server via Docker is the recommended approach, offering significant security and reliability advantages that form the core of the tool’s secure usage pattern:

Security Benefits

Isolation (Primary Mitigation): The Docker container provides essential filesystem and process isolation. AWS CLI commands and piped Unix utilities run in a contained environment. Accidental or misused commands affecting the filesystem are limited to the container, protecting your host machine.
Controlled Credential Access: When mounting credentials, using the :ro (read-only) flag limits the container’s ability to modify your AWS configuration files.
No Local Installation: Avoids installing the AWS CLI and its dependencies directly on your host system.
Clean Environment: Each container run starts with a known, clean state.

Reliability Advantages

Consistent Configuration: All required tools (AWS CLI, SSM plugin, jq) are pre-installed and properly configured
Dependency Management: Avoid version conflicts between tools and dependencies
Cross-Platform Consistency: Works the same way across different operating systems
Complete Environment: Includes all necessary tools for command pipes, filtering, and formatting

Other Benefits

Multi-Architecture Support: Runs on both Intel/AMD (x86_64) and ARM (Apple Silicon, AWS Graviton) processors
Simple Updates: Update to new versions with a single pull command
No Python Environment Conflicts: Avoids potential conflicts with other Python applications on your system
Version Pinning: Easily pin to specific versions for stability in production environments

Versioning

This project uses setuptools_scm to automatically determine versions based on Git tags:

Release versions: When a Git tag exists (e.g., 1.2.3), the version will be exactly that tag
Development versions: For commits without tags, a development version is generated in the format:
<last-tag>.post<commits-since-tag>+g<commit-hash>.d<date> (e.g., 1.2.3.post10+gb697684.d20250406)

The version is automatically included in:

Package version information
Docker image labels
Continuous integration builds

Creating Releases

To create a new release version:

# Create and push a new tag
git tag -a 1.2.3 -m "Release version 1.2.3"
git push origin 1.2.3

The CI/CD pipeline will automatically build and publish Docker images with appropriate version tags.

For more detailed information about the version management system, see VERSION.md.

License

This project is licensed under the MIT License - see the LICENSE file for details.