Introduction to mcp-image-optimizer

This repo includes two Model Context Protocol servers inspired by Green Software Patterns. Rather than another simple weather api MCP server, I wanted to generate two examples that would resonate with a realistic use case.

Green Software Patterns provide guidance to reduce the carbon footprint broken down into three categories:

• Artificial Intelligence (AI)
• Cloud
• Web

Some rules are easier to programmatically enforce than others. Encrypting what is necessary requires business knowledge that's suitable for an agent grounded on architecture documentation and business rules, but doesn't lend itself to a Model Context Protocol Server.

MCP servers extend the native capabilities of the LLM with a REST-like service. The two examples in this repo are inspired by the Serve images in modern formats Green Software pattern.

The goal of the two MCP servers is to search for PNG, JPG, and GIF and convert them to WebP using SixLabors.ImageSharp.

MCP Servers

Mcp.ImageOptimizer.Stdio

Designed to run locally in Visual Studio Code using stdio. This traverses the loaded workspace, identifies image files, and converts image files. The MCP server runs as a subprocess and uses the same identity as the VS Code user. Images are identified using the file extension (PNG, JPG, GIF) and converted to WebP.

stdio Configuration for local testing

This MCP configuration compiles and launches the Mcp.ImageOptimizer.Stdio server. It's useful for debugging.

{
  "servers": {
    "imageoptimizer": {
      "type": "stdio",
      "command": "dotnet",
      "args": [
        "run",
        "-configuration",
        "Debug",
        "--no-build",
        "--project",
        "C:\\Users\\iwasz\\source\\repos\\mcp-image-optimizer\\src\\Mcp.ImageOptimizer.Stdio\\Mcp.ImageOptimizer.Stdio.csproj"
      ]
    }
  }
}

Mcp.ImageOptimizer.StreamingHttp

Uses Kestrel to host a Streamable HTTP MCP server. Uses cached Azure credentials to identify Azure Blob containers. Blob images are identified using the MIME type and converted to WebP. Authentication is managed using an Entra app registration. This is not a recommended approach for a production application. Ideally, OAuth or Entra should be used for authentication (see MCP Entra Examples). This app registration uses a trusted-subsystem model and leads to an overprivileged service.

Streamable HTTP configuration

{
  "servers": {
    "imageoptimizer-azure": {
      "type": "http",
      "monitor": true,

      "url": "http://localhost:5000"
    }
  }
}

Common Projects

Resources