Open source MCP servers for AWS
AWS's official MCP server suite: docs, IaC, containers, serverless, databases, cost and monitoring
Install
{
"mcpServers": {
"awslabs-core-mcp-server": {
"command": "uvx",
"args": ["awslabs.core-mcp-server@latest"],
"env": {
"FASTMCP_LOG_LEVEL": "ERROR"
}
}
}
}macOS and Linux example. Substitute other package names from their READMEs; on Windows use uv tool run.
This is third-party code. Review the repository files before installing.
What it does
The AWS Labs monorepo holds dozens of separate MCP servers, each for its own area. The documentation server gives the agent current AWS references, the IaC server helps with CloudFormation and CDK, and there are servers for EKS, ECS, Lambda and SAM, for DynamoDB, Aurora, Redshift, DocumentDB, Neptune and ElastiCache, and for CloudWatch, CloudTrail, IAM, billing and pricing. All servers run locally over stdio, install from PyPI via uvx or run from images in public ECR. Some servers ship skills, for example for Aurora DSQL and CloudWatch investigations.
Who it is for. For DevOps engineers and developers building and running infrastructure on AWS.
Good fit when
- The agent needs current AWS docs instead of stale model knowledge
- You need to write and validate CloudFormation or CDK
- You need to dig into CloudWatch logs and metrics or account costs
Not a fit when
- Your infrastructure is on other clouds
- You need production agent access with auditing; AWS recommends Agent Toolkit for AWS for that
Example request
Look up current Lambda limits in the AWS docs and validate my CloudFormation templateLimitations
Most servers need AWS credentials with access to the relevant services. Servers support stdio only; SSE was removed. AWS names Agent Toolkit for AWS as the successor, and the most useful projects will move there over time. Some servers are marked deprecated.
How to disable. Remove the server block from your MCP client config and optionally clear the cache with uv cache clean and the server name.
MCP
- Transport
- stdio
- Authentication
- API key
| Environment variables | |
|---|---|
| AWS_REGION | AWS region |
| AWS_ACCESS_KEY_ID secret | AWS access key when not using a profile |
| FASTMCP_LOG_LEVEL | Server log level |
Security check
- Servers operate on real AWS resources and can modify them
- Cloud account credentials are required
README in short
The README opens by noting that Agent Toolkit for AWS is now the successor. It explains how MCP servers improve model answers about AWS and catalogs servers by task with one-click installs for Kiro, Cursor and VS Code. It covers general uv setup, macOS, Linux and Windows configs, container runs and client examples. Apache-2.0 licensed.
FAQ
Which server should I start with?
The AWS documentation server is a good start; add others per service. Each has its own README with options.
Can I run them in Docker?
Yes, images for every server are in public AWS ECR; pass credentials via --env-file.
Related
An MCP server built into the Netdata agent: metrics, logs, alerts and live process, service and container data for an AI assistant
GitHub's official MCP server: code, issues, pull requests, Actions and security alerts straight from the agent
A GitHub Action that runs Claude Code in pull requests and issues: answers mentions, reviews and makes changes
An AI assistant for Kubernetes that turns plain-language requests into kubectl operations and doubles as an MCP server