Content
# Connect AI to Your AWS Resources
Transform how you manage and access your AWS infrastructure by connecting Claude, Cursor AI, and other AI assistants directly to your AWS accounts through AWS IAM Identity Center (formerly AWS SSO). Get instant access to your cloud resources, execute commands, and manage EC2 instances using natural language.
[](https://www.npmjs.com/package/@aashari/mcp-server-aws-sso)
[](https://www.npmjs.com/package/@aashari/mcp-server-aws-sso)
## What You Can Do
✅ **Ask AI about your AWS accounts**: *"Show me all my AWS accounts and available roles"*
✅ **Execute AWS commands**: *"List all S3 buckets in my production account"*
✅ **Manage EC2 instances**: *"Check the disk usage on server i-123456789"*
✅ **Access multi-account setups**: *"Switch to the staging account and describe the VPCs"*
✅ **Monitor resources**: *"Get the status of all running EC2 instances"*
✅ **Run shell commands**: *"Execute 'df -h' on my web server via SSM"*
## Perfect For
- **DevOps Engineers** managing multi-account AWS environments and infrastructure automation
- **Cloud Architects** needing quick access to resource information across AWS accounts
- **Developers** who want to check deployments and run AWS CLI commands through AI
- **SRE Teams** monitoring and troubleshooting AWS resources using natural language
- **IT Administrators** managing EC2 instances and executing remote commands securely
- **Anyone** who wants to interact with AWS using conversational AI
## Quick Start
Get up and running in 2 minutes:
### 1. Get Your AWS SSO Setup
Set up AWS IAM Identity Center:
1. **Enable AWS IAM Identity Center** in your AWS account
2. **Configure your identity source** (AWS directory, Active Directory, or external IdP)
3. **Set up permission sets** and assign users to AWS accounts
4. **Note your AWS SSO Start URL** (e.g., `https://your-company.awsapps.com/start`)
### 2. Try It Instantly
```bash
# Set your AWS SSO configuration
export AWS_SSO_START_URL="https://your-company.awsapps.com/start"
export AWS_REGION="us-east-1"
# Start the authentication flow
npx -y @aashari/mcp-server-aws-sso login
# List your accessible accounts and roles
npx -y @aashari/mcp-server-aws-sso ls-accounts
# Execute an AWS command
npx -y @aashari/mcp-server-aws-sso exec-command \
--account-id 123456789012 \
--role-name ReadOnly \
--command "aws s3 ls"
```
## Connect to AI Assistants
### For Claude Desktop Users
Add this to your Claude configuration file (`~/.claude/claude_desktop_config.json`):
```json
{
"mcpServers": {
"aws-sso": {
"command": "npx",
"args": ["-y", "@aashari/mcp-server-aws-sso"],
"env": {
"AWS_SSO_START_URL": "https://your-company.awsapps.com/start",
"AWS_REGION": "us-east-1"
}
}
}
}
```
Restart Claude Desktop, and you'll see "🔗 aws-sso" in the status bar.
### For Other AI Assistants
Most AI assistants support MCP. Install the server globally:
```bash
npm install -g @aashari/mcp-server-aws-sso
```
Then configure your AI assistant to use the MCP server with STDIO transport.
### Alternative: Configuration File
Create `~/.mcp/configs.json` for system-wide configuration:
```json
{
"aws-sso": {
"environments": {
"AWS_SSO_START_URL": "https://your-company.awsapps.com/start",
"AWS_REGION": "us-east-1",
"DEBUG": "false"
}
}
}
```
**Alternative config keys:** The system also accepts `"@aashari/mcp-server-aws-sso"` or `"mcp-server-aws-sso"` instead of `"aws-sso"`.
## Real-World Examples
### 🔐 Authenticate and Explore
Ask your AI assistant:
- *"Log into AWS SSO and show me my authentication status"*
- *"List all my AWS accounts and the roles I can assume"*
- *"Check if I'm still authenticated to AWS"*
- *"Show me which AWS accounts I have access to"*
### 🛠️ Execute AWS Commands
Ask your AI assistant:
- *"List all S3 buckets in my production account using the ReadOnly role"*
- *"Show me all running EC2 instances in the us-west-2 region"*
- *"Describe the VPCs in my staging AWS account"*
- *"Get the status of my RDS databases in account 123456789012"*
### 🖥️ Manage EC2 Instances
Ask your AI assistant:
- *"Check the disk usage on EC2 instance i-1234567890abcdef0"*
- *"Run 'uptime' on my web server via Systems Manager"*
- *"Execute 'systemctl status nginx' on instance i-abc123 in production"*
- *"Get memory usage from all my application servers"*
### 🔍 Infrastructure Monitoring
Ask your AI assistant:
- *"List all Lambda functions in my development account"*
- *"Show me the CloudFormation stacks in us-east-1"*
- *"Check the health of my load balancers"*
- *"Get the latest CloudWatch alarms that are in ALARM state"*
### 🔄 Multi-Account Operations
Ask your AI assistant:
- *"Switch to account 987654321098 with AdminRole and list all security groups"*
- *"Compare the running instances between staging and production accounts"*
- *"Check backup policies across all my AWS accounts"*
- *"Audit IAM users in the security account"*
<details>
<summary><b>MCP Tool Examples (Click to expand)</b></summary>
### `aws_sso_login`
**Basic Login:**
```json
{}
```
**Custom Login Options:**
```json
{
"launchBrowser": false
}
```
### `aws_sso_status`
**Check Authentication Status:**
```json
{}
```
### `aws_sso_ls_accounts`
**List All Accounts and Roles:**
```json
{}
```
### `aws_sso_exec_command`
**List S3 Buckets:**
```json
{
"accountId": "123456789012",
"roleName": "ReadOnly",
"command": "aws s3 ls"
}
```
**Describe EC2 Instances in a Specific Region:**
```json
{
"accountId": "123456789012",
"roleName": "AdminRole",
"command": "aws ec2 describe-instances --query 'Reservations[*].Instances[*].[InstanceId,State.Name,InstanceType]' --output table",
"region": "us-west-2"
}
```
### `aws_sso_ec2_exec_command`
**Check System Resources:**
```json
{
"instanceId": "i-0a69e80761897dcce",
"accountId": "123456789012",
"roleName": "InfraOps",
"command": "uptime && df -h && free -m"
}
```
</details>
## Transport Modes
This server supports two transport modes for different integration scenarios:
### STDIO Transport (Default for MCP Clients)
- Traditional subprocess communication via stdin/stdout
- Ideal for local AI assistant integrations (Claude Desktop, Cursor AI)
- Uses pipe-based communication for direct MCP protocol exchange
```bash
# Run with STDIO transport (default for AI assistants)
TRANSPORT_MODE=stdio npx @aashari/mcp-server-aws-sso
# Using npm scripts (after installation)
npm run mcp:stdio
```
### HTTP Transport (Default for Server Mode)
- Modern HTTP-based transport with Server-Sent Events (SSE)
- Supports multiple concurrent connections
- Better for web-based integrations and development
- Runs on port 3000 by default (configurable via PORT env var)
- Endpoint: http://localhost:3000/mcp
- Health check: http://localhost:3000/
```bash
# Run with HTTP transport (default when no CLI args)
TRANSPORT_MODE=http npx @aashari/mcp-server-aws-sso
# Using npm scripts (after installation)
npm run mcp:http
# Test with MCP Inspector
npm run mcp:inspect
```
### Environment Variables
**Transport Configuration:**
- `TRANSPORT_MODE`: Set to `stdio` or `http` (default: `http` for server mode, `stdio` for MCP clients)
- `PORT`: HTTP server port (default: 3000)
- `DEBUG`: Enable debug logging (default: false)
**AWS Configuration:**
- `AWS_SSO_START_URL`: Your AWS IAM Identity Center start URL (e.g., `https://your-org.awsapps.com/start`)
- `AWS_SSO_REGION` or `AWS_REGION`: AWS region for SSO authentication (e.g., `us-east-1`)
- `AWS_PROFILE`: AWS profile name (optional, for CLI compatibility)
## Available Tools
When integrated with AI assistants via MCP, the following tools are available:
### Authentication Tools
- **`aws_sso_login`**: Initiates AWS SSO device authorization flow
- Parameters: `launchBrowser` (optional, boolean, default: true)
- Opens browser automatically for authentication
- Handles device authorization code flow
- Caches tokens for subsequent operations
- **`aws_sso_status`**: Checks current authentication status
- No parameters required
- Returns session details and expiration time
- Verifies cached token validity
### Account Management Tools
- **`aws_sso_ls_accounts`**: Lists all accessible AWS accounts and roles
- No parameters required
- Shows account IDs, names, emails, and available roles
- Essential for discovering which accounts/roles you can use
### Command Execution Tools
- **`aws_sso_exec_command`**: Executes AWS CLI commands with SSO credentials
- Required: `accountId`, `roleName`, `command`
- Optional: `region`
- Automatically obtains and caches temporary credentials
- Supports any AWS CLI command
- **`aws_sso_ec2_exec_command`**: Executes shell commands on EC2 instances via SSM
- Required: `instanceId`, `accountId`, `roleName`, `command`
- Optional: `region`
- No SSH access required (uses AWS Systems Manager)
- Instance must have SSM Agent installed
## CLI Commands
All tools are also available as CLI commands using `kebab-case`. Run `--help` for details (e.g., `mcp-aws-sso login --help`).
- **login**: Authenticates via AWS SSO (`--no-launch-browser`). Ex: `mcp-aws-sso login`.
- **status**: Checks authentication status (no options). Ex: `mcp-aws-sso status`.
- **ls-accounts**: Lists accounts/roles (no options). Ex: `mcp-aws-sso ls-accounts`.
- **exec-command**: Runs AWS CLI command (`--account-id`, `--role-name`, `--command`, `--region`). Ex: `mcp-aws-sso exec-command --account-id 123456789012 --role-name ReadOnly --command "aws s3 ls"`.
- **ec2-exec-command**: Runs shell command on EC2 (`--instance-id`, `--account-id`, `--role-name`, `--command`, `--region`). Ex: `mcp-aws-sso ec2-exec-command --instance-id i-0a69e80761897dcce --account-id 123456789012 --role-name InfraOps --command "uptime"`.
<details>
<summary><b>CLI Command Examples (Click to expand)</b></summary>
### Login
**Standard Login (launches browser and polls automatically):**
```bash
mcp-aws-sso login
```
**Login without Browser Launch:**
```bash
mcp-aws-sso login --no-launch-browser
```
### Execute AWS Commands
**List S3 Buckets:**
```bash
mcp-aws-sso exec-command \
--account-id 123456789012 \
--role-name ReadOnly \
--command "aws s3 ls"
```
**List EC2 Instances with Specific Region:**
```bash
mcp-aws-sso exec-command \
--account-id 123456789012 \
--role-name AdminRole \
--region us-west-2 \
--command "aws ec2 describe-instances --output table"
```
### Execute EC2 Commands
**Check System Resources:**
```bash
mcp-aws-sso ec2-exec-command \
--instance-id i-0a69e80761897dcce \
--account-id 123456789012 \
--role-name InfraOps \
--command "uptime && df -h && free -m"
```
</details>
## Troubleshooting
### "Authentication failed" or "Token expired"
1. **Re-authenticate with AWS SSO**:
```bash
# Test your SSO configuration
npx -y @aashari/mcp-server-aws-sso login
```
2. **Check your AWS SSO configuration**:
- Verify your `AWS_SSO_START_URL` is correct (should be your organization's SSO portal)
- Ensure your `AWS_REGION` matches your SSO region configuration
3. **Verify your SSO setup**:
- Make sure you can access the SSO portal in your browser
- Check that your AWS account assignments are active
### "Account not found" or "Role not found"
1. **Check available accounts and roles**:
```bash
# List all accessible accounts
npx -y @aashari/mcp-server-aws-sso ls-accounts
```
2. **Verify account ID format**:
- Account ID should be exactly 12 digits
- Use the exact account ID from the `ls-accounts` output
3. **Check role permissions**:
- Make sure you have permission to assume the specified role
- Use the exact role name from your permission sets
### "AWS CLI not found" or Command execution errors
1. **Install AWS CLI v2**:
- Download from [AWS CLI Installation Guide](https://aws.amazon.com/cli/)
- Ensure `aws` command is in your system PATH
2. **Test AWS CLI independently**:
```bash
aws --version
aws sts get-caller-identity
```
### "EC2 command failed" or "SSM connection issues"
1. **Verify EC2 instance setup**:
- Instance must have SSM Agent installed and running
- Instance needs an IAM role with `AmazonSSMManagedInstanceCore` policy
2. **Check your role permissions**:
- Your assumed role needs `ssm:SendCommand` and `ssm:GetCommandInvocation` permissions
- Verify the instance is in a running state
3. **Test SSM connectivity**:
```bash
# Test if instance is reachable via SSM
npx -y @aashari/mcp-server-aws-sso exec-command \
--account-id YOUR_ACCOUNT \
--role-name YOUR_ROLE \
--command "aws ssm describe-instance-information"
```
### Claude Desktop Integration Issues
1. **Restart Claude Desktop** after updating the config file
2. **Check the status bar** for the "🔗 aws-sso" indicator
3. **Verify config file location**:
- macOS: `~/.claude/claude_desktop_config.json`
- Windows: `%APPDATA%\Claude\claude_desktop_config.json`
### Getting Help
If you're still having issues:
1. Run a simple test command to verify everything works
2. Check the [GitHub Issues](https://github.com/aashari/mcp-server-aws-sso/issues) for similar problems
3. Create a new issue with your error message and setup details
## Frequently Asked Questions
### What permissions do I need?
**For AWS IAM Identity Center (SSO) Setup:**
- Access to AWS IAM Identity Center with a configured identity source
- Permission sets assigned to you by your AWS administrator
- Access to the specific AWS accounts you want to manage
**For EC2 Commands via SSM:**
- Your assumed role needs `ssm:SendCommand` and `ssm:GetCommandInvocation` permissions
- EC2 instances need an IAM role with `AmazonSSMManagedInstanceCore` policy
- SSM Agent must be installed and running on target instances
### Can I use this with multiple AWS organizations?
Currently, each installation supports one AWS SSO start URL at a time. For multiple organizations, you can:
- Switch the `AWS_SSO_START_URL` environment variable between sessions
- Run separate instances with different configurations
- Use multiple Claude Desktop configurations for different organizations
### How long do the SSO credentials last?
- **SSO tokens**: Typically 8-12 hours (managed by AWS IAM Identity Center)
- **Temporary credentials**: Approximately 1 hour per account/role
- The tool automatically handles token refresh and credential caching
- You'll be prompted to re-authenticate when tokens expire
### What AI assistants does this work with?
Any AI assistant that supports the Model Context Protocol (MCP):
- **Claude Desktop** (most popular and well-tested)
- **Cursor AI** (code editor with AI)
- **Continue.dev** (VS Code extension)
- Any other MCP-compatible client
### Is my data secure?
Yes! This tool prioritizes security:
- Runs entirely on your local machine (no external servers)
- Uses your own AWS SSO credentials (no third-party authentication)
- Never sends your data to third parties
- Only accesses what you explicitly grant permission to
- Uses AWS temporary credentials that automatically expire
- Follows AWS best practices for credential management
- Credentials are stored in standard AWS locations (`~/.aws/`)
### Do I need AWS CLI installed?
**For `aws_sso_exec_command`:** Yes, AWS CLI v2 is required to execute AWS commands.
**For other tools:** No, authentication (`aws_sso_login`), status checking (`aws_sso_status`), and account listing (`aws_sso_ls_accounts`) work without AWS CLI.
**For `aws_sso_ec2_exec_command`:** No, this uses the AWS SDK directly via Systems Manager.
### Can I use this with AWS CLI profiles?
This tool uses AWS IAM Identity Center directly and manages its own credential cache. It doesn't require AWS CLI profiles but is compatible with them:
- The tool stores credentials in `~/.aws/sso/cache/` (standard AWS location)
- You can optionally set `AWS_PROFILE` for compatibility with other AWS tools
- The tool works independently of AWS CLI profile configuration
### What's the difference between AWS SSO and AWS IAM Identity Center?
They're the same service! AWS SSO was rebranded to **AWS IAM Identity Center** in 2022. This tool works with both names:
- References to "AWS SSO" in the code and documentation refer to AWS IAM Identity Center
- Your start URL format remains the same: `https://your-org.awsapps.com/start`
- All functionality is identical regardless of the naming
### What is TOON format?
TOON (Token-Oriented Object Notation) is an output format optimized for Large Language Models:
- More compact than JSON (saves tokens when sending data to AI)
- Still human-readable
- Automatically used when available, falls back to JSON if needed
- Learn more: [@toon-format/toon](https://www.npmjs.com/package/@toon-format/toon)
### Where are logs stored?
Debug logs are written to: `~/.mcp/data/@aashari.mcp-server-aws-sso.[session-id].log`
Each session gets a unique log file. Enable debug logging with `DEBUG=true`.
<details>
<summary><b>Response Format Examples (Click to expand)</b></summary>
### Output Format (TOON)
Responses are formatted using **TOON (Token-Oriented Object Notation)** format, which is optimized for LLM token efficiency. TOON provides a more compact representation than JSON while maintaining readability.
**Key Features:**
- Automatically converts responses to TOON format when available
- Falls back to JSON if TOON conversion fails
- Truncates large responses (>10KB) with a note about the full response location
- Logs full responses to `~/.mcp/data/@aashari.mcp-server-aws-sso.[session-id].log`
### MCP Tool Response Example (`aws_sso_exec_command`)
```markdown
# AWS SSO: Command Result
**Account/Role:** 123456789012/ReadOnly
**Region:** us-east-1 (Default: ap-southeast-1)
## Command
aws s3 ls
## Output
2023-01-15 08:42:53 my-bucket-1
2023-05-22 14:18:19 my-bucket-2
2024-02-10 11:05:37 my-logs-bucket
*Executed: 2025-05-19 06:21:49 UTC*
```
### Error Response Example
```markdown
# ❌ AWS SSO: Command Error
**Account/Role:** 123456789012/ReadOnly
**Region:** us-east-1 (Default: ap-southeast-1)
## Command
aws s3api get-object --bucket restricted-bucket --key secret.txt output.txt
## Error: Permission Denied
The role `ReadOnly` does not have permission to execute this command.
## Error Details
An error occurred (AccessDenied) when calling the GetObject operation: Access Denied
### Troubleshooting
#### Available Roles
- AdminAccess
- PowerUserAccess
- S3FullAccess
Try executing the command again using one of the roles listed above that has appropriate permissions.
*Executed: 2025-05-19 06:17:49 UTC*
```
### Large Response Handling
When API responses exceed 10KB, the output is truncated with a message:
```
[Response truncated for AI consumption. Full response logged to: /path/to/log/file.log]
```
This ensures AI assistants receive manageable response sizes while developers can access full output in log files.
</details>
## Technical Details
### Architecture
This server follows a clean 5-layer architecture:
1. **CLI Layer** (`src/cli/`): Command-line interface using Commander.js
2. **Tools Layer** (`src/tools/`): MCP tool definitions with Zod validation schemas
3. **Controllers Layer** (`src/controllers/`): Business logic and orchestration
4. **Services Layer** (`src/services/`): External API interactions (AWS SDK)
5. **Utils Layer** (`src/utils/`): Shared utilities (logging, config, caching, formatting)
### Key Dependencies
- **@modelcontextprotocol/sdk** v1.23.0: MCP protocol implementation
- **@aws-sdk/client-sso** v3.893.0: AWS SSO API client
- **@aws-sdk/client-ssm** v3.893.0: AWS Systems Manager for EC2 commands
- **@toon-format/toon** v2.0.1: Token-efficient output formatting
- **zod** v4.1.13: Runtime type validation
- **commander** v14.0.2: CLI framework
### Logging
Debug logs are written to: `~/.mcp/data/@aashari.mcp-server-aws-sso.[session-id].log`
Enable debug logging by setting `DEBUG=true` in your environment.
### Caching
- **SSO tokens**: Cached in `~/.aws/sso/cache/` (standard AWS location)
- **Temporary credentials**: Cached for 1 hour per account/role combination
- **Account information**: Fetched fresh on each request (no persistent cache)
## Development
```bash
# Clone repository
git clone https://github.com/aashari/mcp-server-aws-sso.git
cd mcp-server-aws-sso
# Install dependencies
npm install
# Build the project
npm run build
# Run in development mode with HTTP transport
npm run dev:http
# Run with STDIO transport (for MCP client testing)
npm run dev:stdio
# Run with MCP Inspector (visual debugging)
npm run mcp:inspect
# Run tests
npm test
# Run tests with coverage
npm test:coverage
# Lint code
npm run lint
# Format code
npm run format
```
### Available npm Scripts
- `npm run build` - Compile TypeScript to JavaScript
- `npm run mcp:stdio` - Run with STDIO transport
- `npm run mcp:http` - Run with HTTP transport
- `npm run mcp:inspect` - Run with MCP Inspector for debugging
- `npm test` - Run Jest tests
- `npm run lint` - Run ESLint
- `npm run format` - Format code with Prettier
## Requirements
- **Node.js**: Version 18.0.0 or higher
- **AWS CLI**: Version 2.x (required only for `aws_sso_exec_command`)
- **AWS IAM Identity Center**: Configured and accessible
- **Operating System**: macOS, Linux, or Windows
## Version History
### v3.0.1 (Current)
- Fixed picomatch dependency conflict for npm ci
- Enhanced raw response logging with truncation for large API responses
- Improved AI guidance for AWS SSO login instructions
### v3.0.0
- **BREAKING**: Modernized to @modelcontextprotocol/sdk v1.23.0 with registerTool API
- Added Node.js version specification (Node 22.14.0 compatibility)
- Enhanced logging and error handling
### v2.0.0
- **BREAKING**: Fixed AWS CLI execution and credential region mismatch issues
- Improved cross-region authentication handling
- Prevented dotenv from outputting to STDIO in MCP mode
See [CHANGELOG.md](./CHANGELOG.md) for complete version history.
## Contributing
Contributions are welcome! Please feel free to submit a Pull Request. For major changes, please open an issue first to discuss what you would like to change.
## License
ISC License - See LICENSE file for details
## Support
Need help? Here's how to get assistance:
1. **Check the troubleshooting section above** - most common issues are covered there
2. **Visit our GitHub repository** for documentation and examples: [github.com/aashari/mcp-server-aws-sso](https://github.com/aashari/mcp-server-aws-sso)
3. **Report issues** at [GitHub Issues](https://github.com/aashari/mcp-server-aws-sso/issues)
4. **Start a discussion** for feature requests or general questions
5. **Check debug logs** at `~/.mcp/data/@aashari.mcp-server-aws-sso.[session-id].log` for detailed error information
---
**Built with:** TypeScript, MCP SDK, AWS SDK for JavaScript v3, TOON Format
*Made with care for DevOps teams who want to bring AI into their AWS workflow.*
