AWS Model Context Protocol (MCP) Server
@alexei-led
A lightweight service that enables AI assistants to execute AWS CLI commands (in safe containerized environment) through the Model Context Protocol (MCP). Bridges Claude, Cursor, and other MCP-aware AI tools with AWS CLI for enhanced cloud infrastructure management.
Overview
What is AWS MCP Server?
It is an MCP server that gives Claude access to all 200+ AWS services by wrapping the AWS CLI rather than each API individually. It lets Claude learn and execute AWS CLI commands on your behalf, with your IAM policy controlling what succeeds. It is for users who want to interact with AWS securely through AI assistants.
How to use AWS MCP Server?
Install prerequisites (AWS CLI, credentials, uv), then configure the server in your MCP client by adding a uvx aws-mcp command to your MCP settings. For stronger isolation, run via Docker with read-only AWS credentials. Claude uses two tools: aws_cli_help to get command documentation and aws_cli_pipeline to execute commands with optional pipes (jq, grep).
Key features of AWS MCP Server
- Provides access to all 200+ AWS services via the AWS CLI
- Two tools: aws_cli_help and aws_cli_pipeline
- Supports stdio and streamable–http transports
- Docker deployment for stronger host isolation
- IAM policy acts as the security boundary
- Configurable timeout, output limit, and sandbox mode
Use cases of AWS MCP Server
- Automating AWS resource management through natural language prompts
- Querying AWS service state and metadata via Claude
- Running AWS CLI commands with piped processing (e.g., filtering with jq)
- Safely delegating AWS operations with fine–grained IAM controls
FAQ from AWS MCP Server
What AWS services can I access?
All 200+ AWS services supported by the AWS CLI are accessible through the two server tools.
How does security work?
Your IAM policy is the security boundary. The server executes whatever AWS commands Claude requests, and IAM controls which commands actually succeed. Use a least–privilege IAM role and never root credentials.
What are the prerequisites?
The AWS CLI installed, AWS credentials configured, and uv installed (for uvx). Optionally, Docker for containerized deployment.
What transports are supported?
The server supports stdio (default) and streamable-http transports. The sse transport is deprecated; use streamable-http for web-based MCP clients.
What configuration options are available?
Key environment variables include AWS_PROFILE (default profile), AWS_REGION (default us-east-1), AWS_MCP_TIMEOUT (300s), AWS_MCP_MAX_OUTPUT (100,000 characters), AWS_MCP_TRANSPORT (stdio), and AWS_MCP_SANDBOX (auto).
