AWS MCP Server
Interact with Amazon Web Services through your AI assistant using the AWS API MCP Server.
AWS MCP Server
This page documents the AWS API MCP Server from awslabs/mcp, published as awslabs.aws-api-mcp-server. It runs AWS CLI commands through the call_aws tool with command validation and optional read-only mode.
The upstream README now marks this server as superseded by the newer AWS MCP Server; it remains available and is documented here.
What it Does
- AWS CLI execution: Runs AWS CLI commands through
call_awswith validation and error handling - Command suggestions: Suggests likely AWS CLI commands for a natural language query
- Read-only mode: Restricts execution to non-mutating operations with
READ_OPERATIONS_ONLY=true - Credential chain: Uses boto3's credential chain or a named AWS profile
Installation
# Using uvx
uvx awslabs.aws-api-mcp-server@latest
Docker:
docker run --rm -i \
-e AWS_REGION=us-east-1 \
-v ~/.aws:/app/.aws \
public.ecr.aws/awslabs-mcp/awslabs/aws-api-mcp-server:latest
Configuration
{
"mcpServers": {
"aws-api": {
"command": "uvx",
"args": ["awslabs.aws-api-mcp-server@latest"],
"env": {
"AWS_REGION": "us-east-1",
"AWS_API_MCP_PROFILE_NAME": "default"
}
}
}
}
AWS_REGION sets the default region. AWS_API_MCP_PROFILE_NAME selects a named AWS profile; if it is omitted, the server follows boto3's default credential chain.
Available Tools
| Tool | Purpose | Category |
|---|---|---|
call_aws | Executes AWS CLI commands with validation and error handling | Commands |
suggest_aws_commands | Suggests AWS CLI commands for a natural language query | Discovery |
get_execution_plan | Provides step-by-step guidance for complex AWS tasks via agent scripts; requires EXPERIMENTAL_AGENT_SCRIPTS=true | Automation |
The full tool descriptions are in the upstream README.
Common Workflows
Security Audit
- Use
call_awsto list IAM roles and inspect their policies - Use
call_awsto list S3 buckets and inspect bucket access settings - Review security group rules for open ports
- Ask for a summary of the findings
Cost Investigation
- Use
call_awsto query AWS Cost Explorer for the current month - Ask for the top-spending services
- Use
call_awsto list EC2 instances and identify idle resources - Generate cost optimization recommendations
Resource Inventory
- Use
call_awswithaws ec2 describe-instancesto list EC2 instances - Use
call_awswithaws s3api list-bucketsto list S3 buckets - Ask for Lambda functions and RDS instances in the target region
- Compile a resource inventory report
Example Prompts
- "List all my EC2 instances"
- "Show me S3 buckets in us-west-2"
- "Create a new security group for web servers"
- "Which IAM roles have overly permissive policies?"
- "What are my top 5 cost centers this month?"
Troubleshooting
| Issue | Likely Cause | Solution |
|---|---|---|
| "Access denied" | IAM credentials lack permissions | Verify the IAM policy allows the requested action |
| Wrong account or profile | The default credential chain selected a different profile | Set AWS_API_MCP_PROFILE_NAME to the intended profile |
| Unexpected region | AWS_REGION is unset or the command targets another region | Set AWS_REGION or pass a region in the command |
| Write operation is rejected | READ_OPERATIONS_ONLY=true is set | Unset it if mutating operations are intended |
| Throttling | AWS API rate limit exceeded | Retry with backoff or request a rate limit increase |
| Deprecation notice | The server has been superseded | Review the AWS MCP Server documentation for the replacement |
Security Notes
- Use least-privilege IAM credentials; IAM permissions remain the primary control
- Set
READ_OPERATIONS_ONLY=truefor read-only exploration - Do not connect the server to untrusted data sources;
call_awsis exposed to prompt injection - Enable CloudTrail for audit logging
Sources
Related Articles & Guides
Accessibility Scanner MCP Server
Accessibility Scanner MCP servers enable AI models to perform WCAG compliance checks, capture annotated screenshots, and generate detailed accessibility reports.
Auth0 MCP Server
Auth0 MCP server enables AI models to interact with Auth0 identity and access management, including applications, APIs, actions, logs, and forms.
Azure DevOps MCP Server
The Azure DevOps MCP Server enables AI models to interact with Azure DevOps, providing capabilities for managing work items, pull requests, pipelines, and more.