← Back to MCP servers

AWS

AWS MCP Server

Interact with AWS services including S3, Lambda, CloudWatch, and ECS from your AI coding assistant.

Token requiredGet credentials →

Last reviewed Mar 2, 2026

Install

claude mcp add aws --transport stdio -- npx -y @aws/aws-mcp-server
Scope: project

Environment variables required:

AWS_ACCESS_KEY_ID=your-access-key-id AWS_SECRET_ACCESS_KEY=your-secret-access-key AWS_REGION=us-east-1

Available tools

ToolDescription
s3_listList S3 buckets and objects
s3_getDownload or read an S3 object
lambda_invokeInvoke an AWS Lambda function
cloudwatch_queryQuery CloudWatch logs and metrics
ecs_list_servicesList ECS services in a cluster

What it does

The AWS MCP Server provides your AI coding assistant with access to core AWS services, enabling it to interact with S3 storage, invoke Lambda functions, query CloudWatch logs, and inspect ECS deployments. This gives the assistant deep infrastructure context when helping you build and debug cloud-native applications.

Having AWS access in your editor is particularly powerful for debugging. The assistant can check CloudWatch logs for errors, inspect S3 contents, test Lambda functions, and verify ECS service health -- all operations that typically require switching between the AWS Console and your code.

Use cases

  • List and read S3 objects to inspect stored data, configs, or artifacts
  • Invoke Lambda functions for testing during development
  • Query CloudWatch logs to debug production errors and performance issues
  • Check ECS service status and health for containerized deployments
  • Explore AWS resources to understand your infrastructure setup

Getting started

  1. Create IAM access credentials at the AWS IAM Console or create a dedicated IAM user with limited permissions.
  2. Set the AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, and AWS_REGION environment variables.
  3. For better security, create an IAM user with only the permissions needed for your workflow. Avoid using root account credentials.
  4. Alternatively, if you use AWS SSO or ~/.aws/credentials profiles, the server can use your existing AWS CLI configuration. Remove the explicit env vars and ensure your default profile is configured.