Semantic Kernel
Microsoft's open-source SDK for integrating large language models into applications with plugins and planners.
Technical explanation
Semantic Kernel is Microsoft's answer to LangChain. It's an SDK for building AI applications, with strong support for C# and Python. The framework came out of Microsoft's own efforts to add AI features to products like Microsoft 365 Copilot. They needed something enterprise-ready, and they open-sourced it.
The architecture centers on plugins and planners. Plugins are collections of functions that the AI can call. These might be native code functions, API calls, or even other AI prompts. Planners take a user's goal and figure out which plugins to invoke and in what order. It's similar to agents in LangChain but with a more structured approach.
Microsoft's tight integration with Azure is both a strength and a constraint. Semantic Kernel works great with Azure OpenAI and Azure AI services. Connectors exist for other providers, but Azure is the first-class citizen. For enterprises already in the Microsoft ecosystem, this makes adoption straightforward. For others, it might feel like vendor lock-in.
The documentation and enterprise features are notably polished. There's strong support for telemetry, security boundaries, and the kind of concerns that matter in corporate environments. If you're building copilot-style features for a .NET application or anything that needs to integrate with Microsoft tools, Semantic Kernel is worth evaluating. For Python-first teams without Azure commitments, LangChain or LlamaIndex might be more natural fits.