AWS MCP Servers
github.com/awslabs/mcp- Category
- Developer Tools
- Rank
- No. 346Tools index
- Pricing
- Free
- Type
- TOOL
- Use case
- Deployment & Operations
- Interfaces
- MCP
- Builder
- awslabs
- GitHub
- 9.7k stars
- Latest release
- 2026.09.20260922000649
- Date
About
AWS's official suite of MCP servers: documentation search, cost analysis, CDK and Terraform guidance, and service-specific servers for Bedrock, DynamoDB, S3 and more.
What it does
It connects AI assistants to AWS context and operations through small, specialized MCP programs. Each client maintains a direct connection to a server, which can fetch remote services or local data and return relevant context or expose workflow tools.
Why it's ranked here
The breadth and official AWS focus make this a useful toolbox for AWS-heavy development. Its standardized client integration and current documentation access are practical strengths. The verdict is tempered by local-only transport support and AWS recommending Agent Toolkit for production agent work, making this suite increasingly transitional.
What's good
The servers divide AWS knowledge and workflows into focused capabilities instead of one undifferentiated integration. They can supply current documentation, specialized service context, and executable workflow tools. Setup guidance spans several popular coding assistants, operating systems, and containers.
Tradeoffs
Current servers support only standard input and output. They do not support Streamable HTTP, and Server Sent Events support was removed from current major versions. Each host maintains a separate connection to each server. AWS also positions Agent Toolkit as the production-oriented successor.
How to use it well
Use it when an MCP-capable coding assistant needs focused AWS context or repeatable cloud workflows during local development. Select only the servers relevant to your task. For production agents needing managed controls, audit visibility, or remote transport, follow AWS guidance toward Agent Toolkit instead.
Technical notes+
README.md defines a stdio-only MCP architecture, documents client setup for Kiro, Cline, Cursor, Windsurf, VS Code, and Claude Code, and lists macOS, Linux, Windows, and container guidance. Python package markers such as src/ecs-mcp-server/awslabs/__init__.py and src/aws-api-mcp-server/awslabs/__init__.py extend the awslabs namespace through pkgutil, while src/billing-cost-management-mcp-server/__init__.py identifies a separate Billing and Cost Management package. Comparable namespace markers appear under the EKS, IAM, RODA, CCAPI, Finch, MSSQL, MySQL, Oracle, Valkey, AWS IaC, and OpenAPI server trees.
Observed
- License
- Apache License 2.0
- Interface
- Model Context Protocol servers using stdio only
- Transport limitation
- Current servers do not support Streamable HTTP or Server Sent Events
- Client setup
- Guidance covers Kiro, Cline, Cursor, Windsurf, VS Code, and Claude Code
- Platform setup
- Documentation covers macOS, Linux, Windows, and containers
- Packaging structure
- Multiple Python server packages share an awslabs namespace
Read from README.md, src/billing-cost-management-mcp-server/__init__.py, src/ecs-mcp-server/awslabs/__init__.py, src/eks-mcp-server/awslabs/__init__.py, src/iam-mcp-server/awslabs/__init__.py, src/roda-mcp-server/awslabs/__init__.py, src/ccapi-mcp-server/awslabs/__init__.py, src/finch-mcp-server/awslabs/__init__.py, src/mssql-mcp-server/awslabs/__init__.py, src/mysql-mcp-server/awslabs/__init__.py, src/oracle-mcp-server/awslabs/__init__.py, src/valkey-mcp-server/awslabs/__init__.py, src/aws-api-mcp-server/awslabs/__init__.py, src/aws-iac-mcp-server/awslabs/__init__.py, src/openapi-mcp-server/awslabs/__init__.py.
Media
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.