AWS Labs MCP Servers

What does this agent do, and when should you use it?

This is AWS Labs' open-source collection of Model Context Protocol servers for clients such as Kiro, Cline, Cursor, Windsurf, Claude Code and VS Code. The servers are organized both by what you are building and by how you work, and span AWS documentation access, infrastructure as code (CloudFormation/CDK), containers (EKS, ECS, Finch), serverless and Lambda, AI services (Bedrock, Kendra, Q Business), databases (DynamoDB, PostgreSQL, MySQL, Aurora DSQL, Neptune, DocumentDB), cost and support workflows, and healthcare scenarios. The servers are designed to support stdio only; Server Sent Events support was removed from all servers on May 26, 2025, and Streamable HTTP is still being worked on. Each server is typically published as an awslabs.* PyPI package launched through uvx, configured with environment variables such as AWS_PROFILE and AWS_REGION. The README also notes that Agent Toolkit for AWS is the successor to these MCP servers, plugins and skills, while this repo continues to work and accept contributions.

Each server is a separate MCP process: a client launches it over stdio (typically uvx awslabs.<server>@latest), and the server then calls the relevant AWS APIs or local tooling. For example, aws-documentation-mcp-server and the remote AWS Knowledge MCP Server fetch current docs and API references; aws-iac-mcp-server provides CloudFormation documentation, CDK best-practice guidance and security validation; eks-mcp-server, ecs-mcp-server and finch-mcp-server handle cluster management, orchestration and local image builds; aws-serverless-mcp-server drives the serverless lifecycle through SAM CLI; lambda-tool-mcp-server exposes Lambda functions as callable tools. Data servers (dynamodb-mcp-server, postgres-mcp-server, mysql-mcp-server, aurora-dsql-mcp-server, documentdb-mcp-server, amazon-neptune-mcp-server and others) run queries and modeling operations directly. Most servers gate write and sensitive-data actions behind explicit flags such as --allow-write, --allow-sensitive-data-access, DDB-MCP-READONLY and ALLOW_WRITE, and return results to the host application as MCP tool responses.

1. Engineers writing CloudFormation or CDK templates in Kiro or Cursor who want the agent to check current AWS docs and validate security.
2. Data engineers who want an AI assistant to inspect DynamoDB models or run SQL against Aurora PostgreSQL.
3. Container platform teams operating EKS clusters, ECS services or local Finch image builds.
4. Developers using SAM CLI for Lambda and serverless apps who want agent help across the lifecycle.
5. Teams connecting enterprise knowledge (Bedrock Knowledge Bases, Kendra, Q Business) to a conversational assistant.
6. Support or operations staff who need to create and track AWS Support cases from an AI client.

How do you install or deploy this agent?

There is no single installer; each server ships as a PyPI package run through uvx, so install uv (which provides uvx) and configure AWS credentials first. For the AWS documentation server the client config looks like this (path and field values should follow the matching README section):

{
  "mcpServers": {
    "awslabs.aws-documentation-mcp-server": {
      "command": "uvx",
      "args": ["awslabs.aws-documentation-mcp-server@latest"],
      "env": {
        "FASTMCP_LOG_LEVEL": "ERROR",
        "AWS_DOCUMENTATION_PARTITION": "aws"
      }
    }
  }
}

Servers with write capability need extra flags and AWS configuration, for example:

{
  "mcpServers": {
    "awslabs.eks-mcp-server": {
      "command": "uvx",
      "args": ["awslabs.eks-mcp-server@latest", "--allow-write", "--allow-sensitive-data-access"],
      "env": { "FASTMCP_LOG_LEVEL": "ERROR" }
    }
  }
}

The README also documents config file locations for Kiro (~/.kiro/settings/mcp.json), Cline, Cursor (.cursor/mcp.json), Windsurf, VS Code (.vscode/mcp.json), Claude Code (.mcp.json) and fx, plus one-click install links for Cursor and VS Code.

How do you use this agent?

Add the server to your client config and restart the client; the agent can then call its tools in conversation. For Cursor, write to .cursor/mcp.json:

{
  "mcpServers": {
    "awslabs.dynamodb-mcp-server": {
      "command": "uvx",
      "args": ["awslabs.dynamodb-mcp-server@latest"],
      "env": {
        "DDB-MCP-READONLY": "true",
        "AWS_PROFILE": "default",
        "AWS_REGION": "us-west-2",
        "FASTMCP_LOG_LEVEL": "ERROR"
      }
    }
  }
}

Database servers also accept connection details on the command line, e.g. PostgreSQL:

uvx awslabs.postgres-mcp-server@latest --connection-string postgresql://[username]:[password]@[host]:[port]/[database]

Start with read-only settings (DDB-MCP-READONLY, ALLOW_WRITE=false) before enabling --allow-write and --allow-sensitive-data-access. For AWS-hosted remote servers such as AWS MCP Server or AWS Knowledge MCP Server, connect through mcp-proxy-for-aws or an HTTP URL:

uvx mcp-proxy-for-aws@latest https://aws-mcp.us-east-1.api.aws/mcp

What are this agent's strengths and limitations?

How does this agent compare with similar options?

Key facts side by side with the most closely related agents.

How does FollowAgents rate this agent?

FAQ