mcp-server-server: Model Context Protocol (MCP) Server Implementation

Overview: What is mcp-server-server?

mcp-server-server is a proof of concept implementation that aims to address the challenges associated with using MCP servers by providing an easier, more flexible solution for AI application development. Traditionally, working with MCP servers has required spawning new processes via stdio, which can be cumbersome and problematic when scaling infrastructure.

🔧 Core Features & MCP Capabilities

This server leverages websockets to expose MCP servers over the network, eliminating the process management overhead on the client side. This allows for more streamlined and scalable interaction between AI applications like Claude Desktop, Continue, Cursor, and others, and their respective data sources or tools. Key features of this implementation include:

1. Ease of Use: By exposing MCP servers via websockets, developers can interact with these servers programatically without the need to manage multiple processes.
2. Scalability: As an AI developer, you can run a single MCP server that exposes various services to different clients, each configured for their specific needs.
3. Flexibility: The implementation is designed to work seamlessly with existing MCP servers and tools, offering a more flexible approach compared to the traditional stdio method.

⚙️ MCP Architecture & Protocol Implementation

The architecture of mcp-server-server is centered around converting an existing MCP server into a network-based one using websockets. This conversion involves wrapping the stdio-based server in a program that connects it over websockets, making it accessible from anywhere via a standard URL.

Step 1: Convert a stdio MCP server to a websocket MCP server

This can be achieved by running:

bun run mcp-server -p <port> -- npx -y @modelcontextprotocol/server-puppeteer@latest

In this command, the mcp-server tool wraps the specified MCP server and binds it to a websockets endpoint.

Step 2: Interact with the MCP server programmatically

Here's an example of how to interact with the MCP server using TypeScript:

import { Client } from "@modelcontextprotocol/sdk/client/index.js";
import { WebSocketClientTransport } from "@modelcontextprotocol/sdk/client/websocket.js";

const transport = new WebSocketClientTransport(new URL("ws://localhost:3001"));

const client = new Client(
  {
    name: "example-client",
    version: "1.0.0",
  },
  {
    capabilities: {},
  }
);
await client.connect(transport);
const tools = await client.listTools();
console.log(
  "Tools:",
  tools.tools.map((t) => t.name)
);
await client.close();

This example demonstrates connecting to an MCP server via websockets and listing available tools.

🚀 Getting Started with Installation

To get started, you need to have Node.js installed. You can install mcp-server using npm or yarn:

npm install -g @modelcontextprotocol/server-wrapper
# Or
yarn global add @modelcontextprotocol/server-wrapper

Next, configure your MCP server by running the mcp-server command with the appropriate arguments.

💡 Key Use Cases in AI Workflows

1. Developing Custom AI Applications: By using mcp-server-server, developers can easily integrate various tools and data sources into their custom applications.
2. Scaling AI Services: With websockets, you can scale your MCP server to handle multiple clients more efficiently.

🔌 Integration with MCP Clients

mcp-server-server supports a range of popular MCP clients, including:

• Claude Desktop: Full support for connecting to various MCP servers and tools.
• Continue: Compatibility matrix is similar to Claude Desktop.
• Cursor: Currently only supports specific tool integrations but can be extended.

See the compatibility matrix below for more details:

MCP Client	Resources	Tools	Prompts	Status
Claude Desktop	✅	✅	✅	Full Support
Continue	✅	✅	✅	Full Support
Cursor	❌	✅	❌	Tools Only

📊 Performance & Compatibility Matrix

mcp-server-server is designed to offer excellent performance, leveraging websockets for efficient data transfer. The protocol flow and data architecture are optimized for handling high volumes of data and requests.

MCP Protocol Flow Diagram

graph TD
    A[AI Application] -->|MCP Client API| B[MCP Server]
    B --> C[MCP Tool]
    style A fill:#e1f5fe
    style C fill:#f3e5f5

Data Architecture Diagram

graph TB
    A[Data Resource] -->|Network| B[Clients/MCP Servers]
    C[Server Configuration] --> D[Clients/MCP Servers]
    I[MCP Client API] --> J[Data Resource]
    E[Data Processing Logic] --> F[Client Responses]
    H[Clients] --> G[mcp-server-server]
    style A fill:#e8f5e8
    style C fill:#e1f5fe
    style D fill:#f3e5f5

🛠️ Advanced Configuration & Security

Advanced configuration and security features include:

• Custom Environment Variables: You can configure environment variables for individual servers to ensure secure access.
• Websocket Endpoint Protection: Secure the websocket endpoint using HTTPS or other authentication mechanisms.

Example of MCP Configuration Code Sample:

{
  "mcpServers": {
    "[server-name]": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-[name]"],
      "env": {
        "API_KEY": "your-api-key"
      }
    }
  }
}

❓ Frequently Asked Questions (FAQ)

1. How do I integrate mcp-server-server with my existing AI applications?

   • You can convert your stdio-based MCP servers into websockets by using the mcp-server command-line tool.

2. Is there documentation for all available tools and their integration details?

   • Yes, detailed integration guides and documentation are available in the official MCP protocol repository.

3. Can I use mcp-server-server with any MCP client?

   • While it is designed to be compatible with most popular clients, some compatibility may vary depending on the specific implementation of the MCP server.

4. How do I ensure security when using websockets for MCP servers?

   • Secure your websocket endpoint using HTTPS and configure environment variables for secure access.

5. What are the performance considerations when running mcp-server-server in production?

   • Performance is optimized, but consider load balancing and scaling strategies to handle high volumes of requests efficiently.

👨‍💻 Development & Contribution Guidelines

Contributors can fork the repository, make their changes, and submit pull requests. Issues and feature requests are welcomed as well. For more information, refer to the CONTRIBUTING.md file in the repository.

🌐 MCP Ecosystem & Resources

The MCP (Model Context Protocol) ecosystem includes a variety of resources for developers building AI applications and integrating MCP servers:

• MCP Documentation: Comprehensive documentation on the protocol and server implementation.
• MCP Community Forum: A platform for discussion and support among community members.
• Official MCP Repositories: Hosting repositories for various tools, servers, and clients.

By contributing to mcp-server-server, you can enhance the capabilities of AI applications and enable more seamless integration across different tools and data sources.