Semantic Kernel for .NET 1.75.0 リリース dotnet-1.75.0

Microsoft has published version 1.75.0 of Semantic Kernel for .NET, the company's open-source SDK for orchestrating large language models in production applications. Semantic Kernel abstracts prompts, plugins, agents, and memory into reusable building blocks, and it underpins a growing number of generative AI integrations both inside Microsoft and across the broader enterprise developer community.

As the version number suggests, 1.75.0 is an incremental minor release rather than a milestone update. While the full release notes should be consulted for specifics, releases at this cadence have typically focused on connector refinements, adjustments to the Agent-related APIs, dependency bumps, and routine bug fixes. No major breaking changes appear to be associated with this drop, which makes it a low-risk update for teams already running on the 1.x line.

Recent Semantic Kernel releases have advanced several strategic threads in parallel, and 1.75.0 likely continues that trajectory. Among the most notable ongoing efforts are the formal integration with the Microsoft Agent Framework, support for the Model Context Protocol (MCP) on both the client and server side, and tighter interoperability with OpenAI's Responses API and Azure AI Foundry. Together these changes have been gradually shifting Semantic Kernel from a prompt-and-plugin toolkit toward a fuller agent runtime.

For .NET developers, the practical appeal remains the SDK's unified abstractions over ChatCompletion, Embedding, and VectorStore providers. The same code paths can target OpenAI, Azure OpenAI, Google, Anthropic, Ollama, ONNX-hosted models, and a growing list of vector databases. This reduces vendor lock-in and makes model swaps — increasingly common as price-performance leaders shift every few months — substantially less invasive than rewriting against a vendor SDK directly.

The broader context is the convergence Microsoft is driving between Semantic Kernel and AutoGen, the Python-first multi-agent framework originally developed by Microsoft Research. The two projects are being unified under the Microsoft Agent Framework banner, and over time their public APIs are expected to converge further, with Semantic Kernel positioned as the production-oriented .NET surface and AutoGen retaining a stronger research and Python orientation. How cleanly that convergence lands across language ecosystems remains to be seen.

Competition in the agent-framework space is intensifying. LangChain and LlamaIndex remain the most visible alternatives in the Python world, while AWS has been promoting its Strands Agents SDK and Google offers the Agent Development Kit (ADK). A common thread across these otherwise divergent stacks is the rapid adoption of MCP as a shared protocol for tool and context exposure, which is gradually turning agent frameworks into interchangeable clients over a common substrate rather than walled gardens. Semantic Kernel's early MCP support fits squarely into that pattern.

For existing users, 1.75.0 is best treated as a standard incremental upgrade. Teams should review the diff in the release notes, pay particular attention to any connector or agent API surface they depend on, and roll forward at their normal cadence. Projects that have not yet adopted the newer agent and MCP-related features may find this a reasonable checkpoint to evaluate them, especially if a migration toward the unified Microsoft Agent Framework is on the medium-term roadmap.

More strategically, the steady release cadence — Semantic Kernel has been shipping minor versions roughly every few weeks — signals that Microsoft intends the .NET SDK to remain a first-class citizen alongside its Python counterpart, despite the gravitational pull of the Python AI ecosystem. For enterprises standardized on .NET, that continuity is arguably as important as any single feature in a given release.