feat: Add MCP (Model Context Protocol) server integration

Implements easy-to-integrate MCP server support using the official C# SDK,
allowing Gemini models to use tools from any MCP server.

Key Features:
- McpTool class implementing IFunctionTool interface
- Auto-discovery of tools from MCP servers
- Support for multiple concurrent MCP servers
- Automatic reconnection and error handling
- Comprehensive configuration options

Implementation:
- McpServerConfig: Configure server connections (command, args, env vars)
- McpToolOptions: Control behavior (auto-reconnect, error handling)
- McpTool: Main class that bridges MCP servers with Google Generative AI
- Async/await pattern with proper IDisposable/IAsyncDisposable support
- Converts MCP tools to Google's FunctionDeclaration format
- Forwards function calls to MCP servers and converts responses

Example Usage:
var config = new McpServerConfig
{
    Name = "my-server",
    Command = "npx",
    Arguments = new List<string> { "-y", "@modelcontextprotocol/server-everything" }
};

using var mcpTool = await McpTool.CreateAsync(config);
var model = new GenerativeModel(apiKey, "gemini-2.0-flash-exp");
model.AddFunctionTool(mcpTool);

Files Added:
- src/GenerativeAI.Tools/Mcp/McpServerConfig.cs
- src/GenerativeAI.Tools/Mcp/McpTool.cs
- src/GenerativeAI.Tools/Mcp/README.md
- samples/McpIntegrationDemo/ (complete example project)
- MCP_INTEGRATION_SUMMARY.md

Dependencies:
- Added ModelContextProtocol v0.4.0-preview.3 NuGet package

MCP_INTEGRATION_SUMMARY.md

@@ -0,0 +1,179 @@
+# MCP Integration Summary
+
+## Overview
+
+This document summarizes the MCP (Model Context Protocol) integration added to the Google_GenerativeAI SDK.
+
+## What Was Added
+
+### 1. Core Implementation (`src/GenerativeAI.Tools/Mcp/`)
+
+#### McpServerConfig.cs
+- `McpServerConfig` class for configuring MCP server connections
+- `McpToolOptions` class for controlling McpTool behavior
+- Support for custom commands, arguments, environment variables, and working directories
+
+#### McpTool.cs
+- Main integration class implementing `IFunctionTool` interface
+- Connects to MCP servers via the official C# SDK
+- Automatically discovers and exposes tools from MCP servers
+- Features:
+  - Auto-discovery of MCP server tools
+  - Automatic reconnection support
+  - Error handling with detailed error responses
+  - Support for multiple concurrent MCP servers
+  - Async/await pattern with proper disposal
+  - Converts MCP tools to Google's `FunctionDeclaration` format
+  - Forwards function calls to MCP servers
+  - Handles MCP responses and converts them to `FunctionResponse`
+
+### 2. Example Project (`samples/McpIntegrationDemo/`)
+
+#### Program.cs
+Comprehensive demo showing:
+- Single MCP server integration
+- Multiple MCP servers running simultaneously
+- Custom configuration with advanced options
+- Real-world usage patterns
+
+#### McpIntegrationDemo.csproj
+- .NET 8.0 project
+- References to GenerativeAI and GenerativeAI.Tools
+
+### 3. Documentation
+
+#### samples/McpIntegrationDemo/README.md
+- Quick start guide
+- Usage examples (basic, multiple servers, custom config)
+- Configuration options reference
+- List of available MCP servers
+- API reference
+- Troubleshooting guide
+- Advanced usage patterns
+
+#### src/GenerativeAI.Tools/Mcp/README.md
+- Comprehensive technical documentation
+- Architecture overview
+- Core components explanation
+- Detailed usage examples
+- Error handling patterns
+- Best practices
+- Limitations and troubleshooting
+
+### 4. Dependencies
+
+#### Updated GenerativeAI.Tools.csproj
+- Added `ModelContextProtocol` NuGet package (v0.4.0-preview.3)
+- Official C# SDK for Model Context Protocol
+
+## Key Features
+
+1. **Easy Integration**: Simple API to connect any MCP server
+2. **Auto-Discovery**: Automatically finds and exposes all tools
+3. **Multiple Servers**: Connect to multiple MCP servers at once
+4. **Robust**: Auto-reconnect, error handling, timeouts
+5. **Type-Safe**: Fully typed with nullable reference types
+6. **Well-Documented**: Comprehensive docs and examples
+7. **IDisposable**: Proper resource management
+8. **Async**: Full async/await support
+
+## Usage Example
+
+```csharp
+// Configure MCP server
+var config = new McpServerConfig
+{
+    Name = "my-server",
+    Command = "npx",
+    Arguments = new List<string> { "-y", "@modelcontextprotocol/server-everything" }
+};
+
+// Create MCP tool
+using var mcpTool = await McpTool.CreateAsync(config);
+
+// Add to Gemini model
+var model = new GenerativeModel(apiKey, "gemini-2.0-flash-exp");
+model.AddFunctionTool(mcpTool);
+model.FunctionCallingBehaviour.AutoCallFunction = true;
+
+// Use it!
+var result = await model.GenerateContentAsync("Use the echo tool to say hello!");
+```
+
+## Architecture
+
+```
+Gemini Model (GenerativeAI)
+    ↓ IFunctionTool interface
+McpTool (GenerativeAI.Tools.Mcp)
+    ↓ MCP Protocol
+McpClient (ModelContextProtocol SDK)
+    ↓ Stdio Transport
+MCP Server (Node.js, Python, C#, etc.)
+```
+
+## Files Added
+
+```
+src/GenerativeAI.Tools/
+├── GenerativeAI.Tools.csproj (modified - added ModelContextProtocol package)
+└── Mcp/
+    ├── McpServerConfig.cs (new)
+    ├── McpTool.cs (new)
+    └── README.md (new)
+
+samples/McpIntegrationDemo/
+├── McpIntegrationDemo.csproj (new)
+├── Program.cs (new)
+└── README.md (new)
+
+MCP_INTEGRATION_SUMMARY.md (new)
+```
+
+## Compatibility
+
+- **.NET Target Frameworks**: net8.0, netstandard2.0, net9.0, net462
+- **MCP SDK**: ModelContextProtocol v0.4.0-preview.3
+- **MCP Servers**: Any server implementing the Model Context Protocol specification
+
+## Testing
+
+The integration should be tested with:
+
+1. **Unit Tests**: Test McpTool creation, connection, and function calling
+2. **Integration Tests**: Test with actual MCP servers
+3. **Examples**: Run the demo project with various MCP servers
+
+To run the demo:
+```bash
+export GEMINI_API_KEY=your-api-key
+cd samples/McpIntegrationDemo
+dotnet run
+```
+
+## Future Enhancements
+
+Potential improvements for future versions:
+
+1. **HTTP Transport**: Support HTTP-based MCP servers (requires SDK support)
+2. **Connection Pooling**: Reuse connections for better performance
+3. **Caching**: Cache tool definitions to reduce refresh overhead
+4. **Metrics**: Add telemetry and monitoring
+5. **Testing**: Unit and integration tests
+6. **More Examples**: Additional real-world examples
+7. **Configuration**: JSON-based configuration file support
+
+## References
+
+- [MCP Official Website](https://modelcontextprotocol.io/)
+- [MCP C# SDK](https://github.com/modelcontextprotocol/csharp-sdk)
+- [MCP Specification](https://spec.modelcontextprotocol.io/)
+- [MCP Servers Repository](https://github.com/modelcontextprotocol/servers)
+
+## License
+
+This integration follows the same license as the Google_GenerativeAI SDK.
+
+## Contributing
+
+Contributions, bug reports, and feature requests are welcome!

samples/McpIntegrationDemo/McpIntegrationDemo.csproj

@@ -0,0 +1,16 @@
+<Project Sdk="Microsoft.NET.Sdk">
+
+    <PropertyGroup>
+        <OutputType>Exe</OutputType>
+        <TargetFramework>net8.0</TargetFramework>
+        <ImplicitUsings>enable</ImplicitUsings>
+        <Nullable>enable</Nullable>
+        <LangVersion>latest</LangVersion>
+    </PropertyGroup>
+
+    <ItemGroup>
+      <ProjectReference Include="..\..\src\GenerativeAI\GenerativeAI.csproj" />
+      <ProjectReference Include="..\..\src\GenerativeAI.Tools\GenerativeAI.Tools.csproj" />
+    </ItemGroup>
+
+</Project>