mcp-proxy-for-aws

MCP Proxy for AWS provides a lightweight bridge that signs MCP requests with AWS SigV4 credentials, allowing MCP clients and AI agents to connect to MCP servers hosted on AWS that require IAM authentication. You can use it either as a proxy, sitting between your MCP client (such as Claude Desktop or Kiro CLI) and the AWS-hosted MCP server, or as a library to programmatically connect frameworks like LangChain, LlamaIndex, and Strands Agents to AWS MCP servers. When run as a proxy, install and run the service locally to automatically sign requests; when used as a library, import and instantiate the proxy within your Python applications to handle session management and request signing for you. The tool handles discovery of available tools dynamically and signs each request with your local AWS credentials, negating the need to implement SigV4 signing in your client code.

Prerequisites:

• Python 3.10+
• uv package manager (uvx) installed
• AWS credentials configured (via AWS CLI, environment variables, or IAM roles)
• Optional: Docker (if you plan to run a containerized setup)

Install and run via PyPI (recommended):

# Install and run the MCP Proxy for AWS using uvx
uvx mcp-proxy-for-aws@latest <SigV4 MCP endpoint URL>

Run from a local repository (alternative):

# Clone the repository and run locally
git clone https://github.com/aws/mcp-proxy-for-aws.git
cd mcp-proxy-for-aws
uv run mcp_proxy_for_aws/server.py <SigV4 MCP endpoint URL>

Docker (optional):

# Pull the pre-built image
docker pull public.ecr.aws/mcp-proxy-for-aws/mcp-proxy-for-aws:latest

# Run the container
docker run -e ENDPOINT="<SigV4 MCP endpoint URL>" -it public.ecr.aws/mcp-proxy-for-aws/mcp-proxy-for-aws:latest

Tips and common considerations:

• Ensure your AWS credentials are available to the running process (environment variables, AWS CLI config, or IAM roles).
• The proxy signs requests automatically; no changes are typically required in MCP clients.
• If you encounter startup delays on first run, it may be due to dependency caching when using uvx; subsequent launches should be faster.
• When using Docker, expose necessary environment variables for region/profile if your AWS setup requires them.
• The tool can discover tools dynamically; you can enable or disable write-enabled tools with the appropriate flags (see --read-only option in documentation).
• If you use a Docker image, you can pin to a specific version (e.g., 1.1.6) for reproducible deployments.