Semantic Kernel + Playwright MCP Server Demo

🚀 Welcome to the MCP + Semantic Kernel Demo App!

This project demonstrates how to combine Microsoft Semantic Kernel with the Model Context Protocol (MCP) server using Playwright to enable AI-driven browsing and summarization capabilities.

It uses:

• Azure OpenAI (GPT-4)
• Semantic Kernel Function Calling
• Playwright MCP Server
• Automatic browser interaction
• Bing News summarization using AI

🔧 Features

• Uses Semantic Kernel to interact with AI functions
• Leverages Playwright MCP server to simulate browser-like capabilities
• Demonstrates automatic function calling
• Summarizes AI-related news from Bing by navigating and extracting content

🧠 Prerequisites

Ensure you have the following:

• .NET 8.0 SDK or newer
• Azure OpenAI resource (with deployed model like gpt-4-1106-preview)
• Access to Bing News (public)
• Playwright installed via npx
• MCP Server libraries (ModelContextProtocol)

📦 Setup

1. Clone the Repository

   git clone https://github.com/yourusername/semantic-kernel-playwright-mcp.git
   cd semantic-kernel-playwright-mcp
   

2. Install Dependencies Make sure to restore NuGet packages.

3. Configure Azure OpenAI Replace the following placeholders in Program.cs:

   builder.AddAzureOpenAIChatCompletion(
       "GPT4ov1", 
       "https://<replace>.openai.azure.com", 
       "<replacewithkey>");
   

4. Ensure Playwright MCP is Installed

   npx -y @playwright/mcp@latest
   

▶️ Run the App

dotnet run

You'll see output like:

Welcome to My MCP + Semantic Kernel Demo App!
Running application logic...

Summarize AI news for me related to MCP on bing news. Open first link and summarize content
> <summarized result here>

▶️ Output

🧠 How It Works

1. Initializes the Semantic Kernel with Azure OpenAI.
2. Starts a Playwright MCP Server in stdio mode.
3. Maps MCP actions (like open page, extract content) to SK functions.
4. Invokes a natural language prompt to fetch and summarize news content.
5. Uses automatic function calling to delegate the task to the browser.

🛠 Code Highlights

• GetMCPClientForPlaywright() creates an MCP client using Playwright.
• MapToFunctionsAsync() integrates MCP capabilities into Semantic Kernel.
• Prompts like "Summarize AI news..." trigger both browsing and summarization.

🧪 Example Use Case

You can modify the prompt to:

var prompt = "Search GitHub for latest Semantic Kernel issues and summarize the top 3.";

📁 Project Structure

├── Program.cs
├── AzureAIAgentServiceDemo.csproj
├── README.md
└── ...

📌 Future Ideas

• Add UI using Blazor or Console enhancements
• Integrate multi-turn conversations using AgentGroupChat
• Extend MCP to handle screenshots or downloads

🤝 Contributing

Feel free to fork, enhance, or suggest improvements via pull requests!

📜 License

MIT License. See LICENSE file for details.

Let me know if you'd like to include diagrams or badge-style highlights too!