AWS Labs' official server suite — current AWS docs, CDK guidance, cost analysis and service tools.
AWS documentation is vast, constantly changing, and exactly the kind of thing LLMs confidently misremember. AWS Labs' MCP server suite attacks that problem at the source: rather than one monolithic server, it's a family of focused servers — documentation search, CDK guidance, cost analysis, Bedrock knowledge bases, and service-specific tools for DynamoDB, S3, Lambda and more — each installable independently via uvx.
The documentation server is the load-bearing one. It searches and reads current AWS docs, so "what's the cold-start behaviour of Lambda SnapStart for Python?" gets answered from today's documentation instead of a two-year-old memory of it. For a platform that ships hundreds of changes a month, that recency is the difference between guidance and folklore.
The CDK and Terraform servers add prescriptive infrastructure-as-code patterns with security checks (cdk-nag integration) — ask for "a serverless API with DynamoDB and proper IAM" and the guidance follows AWS's own well-architected patterns. The cost server explains and forecasts spend, which pairs dangerously well with the honesty of an assistant that reads your actual bill.
Being AWS, credentials follow the standard chain (profiles, IAM roles, SSO) — the servers use whatever access your environment already has, so scope with IAM as you would any tool. Start with the docs server (zero AWS credentials needed) and add service servers as workflows demand. For anyone building on AWS with an assistant, the docs server alone eliminates a whole genre of hallucinated-API pain.
AWS_PROFILE is read per server entry, so one assistant can hold docs plus read-only DynamoDB and nothing more.@latest in the uvx args — AWS Labs ships to this repo constantly, and new service-specific servers appear there faster than any list of them stays accurate.Compare MCP servers by category, or browse hosted MCP servers if you prefer an endpoint over a local process.
Service behaviour, quotas and APIs answered from current docs, not memory.
CDK and Terraform patterns that follow AWS's own architectural guidance.
"Why did costs jump last month?" answered from real billing data.
Check the linked publisher documentation and your connected client's tool list. Names and availability can vary with the server version and account permissions.
| Tool | What it does |
|---|---|
| search_documentation | Search current AWS documentation |
| read_documentation | Read a docs page as markdown |
| recommend | Related documentation recommendations |
| (suite) CDK / Terraform guidance | IaC patterns with security checks, via dedicated servers |
{
"mcpServers": {
"aws-docs": {
"command": "uvx",
"args": ["awslabs.aws-documentation-mcp-server@latest"]
}
}
}The docs server needs no AWS credentials. Service servers (DynamoDB, S3…) use your standard AWS credential chain.
Python with uv. Docs server: no credentials. Service/cost servers: standard AWS credential chain with IAM scoping.
| Variable | Description | Required |
|---|---|---|
| AWS_PROFILE | AWS profile for credentialed service servers | Optional |
| AWS_REGION | Default region for service operationse.g. us-east-1 | Optional |
Stop letting your assistant hallucinate n8n node parameters — this server hands it the real schemas, templates and validation.
Cloud browsers for AI agents — automation sessions that run in Browserbase's fleet, not on your machine.
A Model Context Protocol (MCP) server that transforms natural language questions into optimized KQL queries with intelligent schema discovery
[](https://www.npmjs.com/package/@aws-sdk/middleware-expect
Kubernetes RCA agent that sandboxes its own AI with OPA Gatekeeper. MCP-ready for any client.
Open-source infrastructure for AI agents to discover youth nonprofits and send them money directly. No intermediary. No platform fees. Nonprofits keep 100%.