AWS MCP

What is AWS MCP?

AWS MCP is a Model Context Protocol server that enables AI assistants like Claude to interact with your AWS environment using natural language. It allows querying and managing AWS resources during conversations, serving as a local alternative to Amazon Q.

How to use AWS MCP?

Clone the repository, install dependencies with npm or pnpm, then configure Claude Desktop by editing its config file to point to the project. Start by selecting an AWS profile or directly asking natural language queries such as listing EC2 instances or S3 buckets.

Key features of AWS MCP

• Query and modify AWS resources via natural language
• Support for multiple AWS profiles and SSO authentication
• Multi-region support
• Secure local credential handling
• Runs entirely on your machine

Use cases of AWS MCP

• List available AWS profiles interactively
• Query all EC2 instances in an account
• Show S3 buckets with their sizes
• List Lambda functions deployed in a specific region
• List ECS clusters and their services

FAQ from AWS MCP

What prerequisites are needed?

Node.js, Claude Desktop, and locally configured AWS credentials in ~/.aws/.

How do I configure the server in Claude Desktop?

Edit claude_desktop_config.json to add the MCP server entry with the correct path to the project and run command.

Does the server expose my AWS credentials?

No – credentials are used only locally and are never sent to external services.

How can I see logs if something goes wrong?

Tail the log files at ~/Library/Logs/Claude/mcp-server-aws.log or ~/Library/Logs/Claude/mcp.log.

Does the server support multiple AWS regions?

Yes, it includes multi-region support for querying resources across regions.