aws/mcp-proxy-for-aws

What it solves

This project provides a bridge for AI applications and agents to connect to Model Context Protocol (MCP) servers hosted on AWS that require AWS IAM authentication (SigV4). Since standard MCP clients (like Claude Desktop) do not natively support SigV4 signing, this tool handles the authentication process automatically using local AWS credentials.

How it works

It operates in two primary modes:

1. As a Proxy: It acts as a lightweight client-side bridge. An AI assistant connects to the proxy, and the proxy signs the requests with AWS SigV4 before forwarding them to the AWS-hosted MCP server.
2. As a Library: It provides authenticated transport layers that can be integrated directly into Python-based AI agent frameworks (such as LangChain, LlamaIndex, and Strands Agents) to handle the session lifecycle and authentication programmatically.

Who it’s for

• Developers using AI assistants (e.g., Claude Desktop, Kiro CLI) that need to access AWS-secured MCP servers.
• AI engineers building agents with Python frameworks who want to avoid writing custom SigV4 request signing logic.

Highlights

• Automatic SigV4 Signing: Uses local AWS credentials from the CLI, environment variables, or IAM roles.
• Multi-account Access: Supports per-call AWS profile switching, allowing agents to work across multiple AWS accounts without restarting.
• Flexible Deployment: Available as a version-pinned CLI tool, a Python library, or a Docker image.
• Framework Integration: Ready-made patterns for integration with LangChain, LlamaIndex, and Microsoft Agent Framework.