MCPSharp is a .NET library for building Model Context Protocol (MCP) servers and clients, enabling seamless integration between AI models and .NET applications. It simplifies exposing .NET methods as AI-discoverable tools and connecting to existing MCP services, targeting developers who want to add AI capabilities to their applications or create MCP-compliant tools.

How It Works

MCPSharp leverages a .NET attribute-based system ([McpTool], [McpParameter]) to define and expose methods as MCP endpoints. It handles the underlying JSON-RPC communication and parameter validation automatically. The library also supports dynamic tool registration and integrates with Microsoft.Extensions.AI and Semantic Kernel, allowing for flexible tool management and interoperability.

Quick Start & Requirements

• Install: dotnet add package MCPSharp
• Prerequisites: .NET Standard 2.0 or later. XML documentation generation is recommended for rich tool descriptions.
• Quick Start: https://github.com/afrise/MCPSharp#quick-start
• API Reference: https://github.com/afrise/MCPSharp#api-reference

Highlighted Details

• Integrates with Microsoft.Extensions.AI for AIFunctions and Semantic Kernel for KernelFunctionAttribute.
• Supports dynamic tool registration with custom implementation logic.
• Provides tool change notifications to clients when tools are added or updated.
• Handles complex object parameters and offers improved error handling with detailed stack traces.
• Automatically extracts tool descriptions from XML comments.

Maintenance & Community

• The project welcomes contributions via Pull Requests.
• No specific community channels (Discord/Slack) or roadmap links are provided in the README.

Licensing & Compatibility

• Licensed under the MIT License.
• Permissive license suitable for commercial use and integration with closed-source applications.

Limitations & Caveats

The README indicates that [McpFunction] is deprecated in favor of [McpTool], and MCPServer.RegisterTool<T>() is deprecated in favor of MCPServer.Register<T>(). While the old methods still work, migration is recommended for consistency.