Microsoft Advances AI Integration with Model Context Protocol (MCP) C# SDK 1.0

Microsoft has released version 1.0 of its official Model Context Protocol (MCP) C# SDK, fully supporting the November 25, 2025, version of the MCP Specification. This release provides developers with enhanced capabilities for building MCP servers and clients within the .NET ecosystem, streamlining the integration of AI applications with external data sources and tools.

Enhanced Authorization Server Discovery

A key improvement in the new SDK is enhanced authorization server discovery. The updated specification allows servers to expose Protected Resource Metadata through three distinct methods, offering greater flexibility compared to the previous single-method requirement. The SDK automates the full discovery process on the client side, simplifying implementation for developers.

Icon Support for Improved User Experience

Developers can now enrich the user experience by incorporating icons for tools, resources, and prompts. This is achieved through a simple attribute parameter or more advanced configurations, including support for multiple icons, MIME types, size hints, and theme preferences. Server and client metadata also benefit from this enhancement, allowing for the inclusion of icons and website URLs.

Incremental Scope Consent: Prioritizing Security

The release introduces incremental scope consent, a security-focused feature that applies the principle of least privilege to MCP authorization. Instead of requiring clients to request all possible permissions upfront, the new mechanism allows them to start with minimal scopes and request additional access as needed. The SDK manages this process automatically on the client side, reducing complexity and enhancing security.

URL Mode Elicitation for Sensitive Interactions

To further protect sensitive data, the SDK now supports URL mode elicitation. This feature enables servers to redirect sensitive interactions, such as API key collection or third-party authorization, away from the MCP client and to a secure, server-hosted URL. This prevents sensitive information from being transmitted through the client, mitigating potential security risks.

Tool Calling Support in Sampling

One of the most technically significant additions is tool calling support in sampling. Servers can now include tools in sampling requests, enabling a language model to invoke those tools during response generation. Integration with the Microsoft.Extensions.AI package simplifies implementation for both client and server sides.

Additional Improvements

The release also includes several other enhancements:

• OAuth Client ID Metadata Documents (CIMDs) are now preferred over Dynamic Client Registration.
• Improved support for long-running requests over HTTP through a polling mechanism, allowing servers to close connections and clients to reconnect using event IDs.
• Experimental Tasks support for durable state tracking.
• `DistributedCacheEventStreamStore` for Server-Sent Events (SSE) stream storage.
• Extended `Implementation` metadata properties.

Resources for Developers

Detailed implementation steps and code examples for each feature are available in the official .NET Developer Blog. The C# SDK repository is hosted on GitHub, providing access to the source code and release notes.

The Model Context Protocol (MCP) is an open protocol designed to standardize integrations between AI apps and external tools and data sources, as outlined by Microsoft Learn. It follows a client-server architecture, with MCP Hosts initiating connections, MCP Clients maintaining those connections, and MCP Servers providing context and capabilities (GitHub). MCP aims to enable more accurate, relevant, and context-aware responses from AI models (modelcontextprotocol.io).