Semantic Kernel (28,422 Stars) 2026: Microsoft's SDK for Building AI Agents with Plugins and Memory

Semantic Kernel (28,422 stars) is Microsoft's open-source SDK for building AI agents and copilots with plugins, planners and memory. Here is how to get started in C# or Python.

💡 What You Will Learn

Semantic Kernel (28,422 stars) is Microsoft's open-source SDK for building AI agents and copilots with plugins, planners and memory. Here is how to get started in C# or Python.

## The short answer **microsoft/semantic-kernel** (28,422 stars, C#) is Microsoft's open-source SDK that lets you orchestrate AI agents in your applications. It provides plugins (skills), planners, memory, and connectors for models like OpenAI and Azure OpenAI - with first-class support in C#, Python, and Java. ## Core concepts | Concept | What it does | |---|---| | **Kernel** | Central orchestrator that connects models, memory and plugins | | **Plugin** | A group of functions the AI can call | | **Planner** | Automatically sequences plugin calls to fulfill a goal | | **Memory** | Stores and retrieves facts/embeddings for context | | **Connector** | Bridges to model providers (OpenAI, Azure, local) | ## Minimal example (C#) ```csharp using Microsoft.SemanticKernel; var builder = Kernel.CreateBuilder(); builder.AddOpenAIChatCompletion("gpt-4o-mini", apiKey); var kernel = builder.Build(); var result = await kernel.InvokePromptAsync( "Summarize this in one sentence: {{$input}}", new() { ["input"] = "..." }); Console.WriteLine(result); ``` Python version uses the same concepts: ```python from semantic_kernel import Kernel kernel = Kernel() kernel.add_service(OpenAIChatCompletion("gpt-4o-mini", api_key)) ``` ## When to use Semantic Kernel vs LangChain - Semantic Kernel shines when you live in the Microsoft/.NET ecosystem (C#, Azure). - It is designed for enterprise copilots with structured plugins and governance. - LangChain (143,505 stars) is broader with more community integrations; SK is more opinionated and Microsoft-centric. ## Practical tips - Start with plugins over planners - explicit function calls are easier to debug. - Use the memory connectors (vector stores) to give your agent long-term context. - Take advantage of the auto-function-calling to let models invoke your plugins safely. ## FAQ **Is it free?** Yes - MIT licensed open source. **Which languages?** C#, Python, and Java are officially supported. **Does it work with local models?** Yes - you can connect any OpenAI-compatible endpoint, including Ollama.