AWS Labs MCP
Summary
AWS Labs' core MCP server — the entry point to AWS's official collection of service-specific MCP servers.
Features
- Entry point to the AWS Labs MCP server collection
- Routes requests to specialised AWS service servers
- Covers documentation, CDK, Lambda, and more
Installation
Set up this MCP server in your favorite AI agent — copy a ready-made configuration below.
Description
AWS Labs maintains a large collection of MCP servers covering individual AWS services, and the Core server is the front door: it helps an assistant discover and route to the right one.
The collection
Rather than one enormous server, AWS publishes focused servers for documentation lookup, CDK and CloudFormation authoring, Lambda, DynamoDB, S3, Bedrock knowledge bases, cost analysis, ECS and EKS, Aurora, and many more. Each is installed independently and does one thing well.
What Core adds
It provides guidance on which server handles a given task and how to configure it, which matters because the collection is large enough that finding the right server is itself a question.
Why the AWS documentation server deserves special mention
AWS's API surface is vast and changes constantly, and models reliably generate plausible-but-wrong AWS code. A documentation server that fetches current reference material substantially improves the output — often more than any other single addition.
Installed with uvx awslabs.core-mcp-server@latest, with each specialised server added as needed. Uses your existing AWS credentials.
Related MCP Servers
DigitalOcean
by DigitalOcean
DigitalOcean's official MCP server — deploy apps and manage Droplets, databases, and storage conversationally.
Heroku
by Heroku
Heroku's official MCP server — manage apps, dynos, add-ons, config, and logs from an AI client.
Azure
by Microsoft
Microsoft's official Azure MCP server — manage cloud resources and query monitoring data conversationally.

Kubernetes MCP Server
by containers (Red Hat)
Native Go MCP server for Kubernetes and OpenShift — talks to the API server directly, no kubectl required.