Infrastructure for Seamless AI Connectivity

Introducing an easier, more secure way to run MCP

We are happy to announce the release of open-source client / server Python packages that enable secure, distributed hosting and use of MCP tools. Featuring a new SSH-based transport layer for encrypted communication and authentication.

WHAT WE’RE RELEASING

1. m2m-mcp-server-ssh-server (GitHub) - An SSH server that:

   • Hosts and aggregates multiple MCP servers under a unified interface

   • Provides secure key-based authentication

   • Includes an optional HTTP API for key management

   • Lets you proxy any MCP-compliant tool

2. m2m-mcp-server-ssh-client (GitHub) - An SSH client and MCP (proxy) server that:

   • Connects to remote MCP servers via SSH tunneling

   • Proxies local MCP protocol requests to remote servers

   • Supports automatic key exchange for simplified authentication

   • Integrates easily with Claude Desktop and other MCP hosts

3. Public Demo Server: We’ve deployed a ready-to-use demo server at mcp-beta.machinetomachine.ai with 3 MCP servers pre-installed (HackerNews, MLB Stats API, Formula 1 API).

TECHNICAL ADVANTAGES

Our SSH-based implementation offers distinct advantages for distributed workflows:

• Returns full control to users - Eliminates dependency on centralized authentication systems

• Offers flexible deployment - Deploy on AWS/GCP/Azure, on-premises servers, or edge devices like Raspberry Pi

• Provides end-to-end encryption - Built on SSH's proven encryption standard

• Supports server aggregation - Combine multiple MCP servers behind a single interface

• Scales horizontally - Run compute-intensive tools on powerful hardware, access from anywhere

QUICK START EXAMPLE

Connect Claude Desktop to our demo server by adding this to your settings:

"mcpServers": {
  "remote-mcp-tools": {
    "command": "uvx",
    "args": [
      "m2m-mcp-server-ssh-client", 
      "--host", "mcp-beta.machinetomachine.ai", 
      "--port", "8022", 
      "--use-key-server"
    ]
  }
}

Or for CLI testing:

# Install
uv add m2m-mcp-server-ssh-client

# Connect to demo server
uvx m2m-mcp-server-ssh-client --host mcp-beta.machinetomachine.ai --use-key-server

# Debug with MCP Inspector
npx @modelcontextprotocol/inspector -- uvx m2m-mcp-server-ssh-client --host mcp-beta.machinetomachine.ai --use-key-server

For detailed deployment instructions, check the cloud deployment guide.

WHY IT MATTERS

This implementation gives you complete control over your MCP infrastructure. Run tools on specialized hardware, maintain regulatory compliance with on-premise solutions, or build sophisticated multi-server architectures — all while keeping the interface simple for clients.

We’re committed to fostering an open AI ecosystem where tools can be freely developed, hosted, shared, and collaboratively improved, independent of any single provider.

We welcome your feedback and contributions to both packages!