explaingit

microsoft/semantic-kernel

📈 Trending★ 27,844C#Audience · developerComplexity · 3/5ActiveLicenseSetup · moderate

TLDR

Microsoft SDK for building AI agents that connect your code to language models, with plugins for real-world tasks and multi-agent orchestration.

Mindmap

mindmap
  root((Semantic Kernel))
    What it does
      Connect code to LLMs
      Build AI agents
      Multi-agent workflows
    Key features
      Plugins for tools
      Vector database support
      Model-agnostic
    Tech stack
      Python
      .NET/C#
      Java
    Use cases
      Chatbots
      Automated assistants
      Enterprise workflows
    Audience
      Enterprise developers
      AI application builders

Things people build with this

USE CASE 1

Build chatbots that answer questions by connecting to your company's data and APIs.

USE CASE 2

Create automated assistants that handle multi-step workflows by coordinating multiple specialized agents.

USE CASE 3

Add AI-powered search to applications using vector databases to find relevant documents by meaning.

USE CASE 4

Develop enterprise applications that swap between different AI providers without rewriting code.

Tech stack

C#.NETPythonJavaOpenAIAzure OpenAI

Getting it running

Difficulty · moderate Time to first run · 30min

Requires an OpenAI or Azure OpenAI API key to run meaningful examples.

Use freely for any purpose including commercial, as long as you keep the copyright notice.

In plain English

Semantic Kernel is Microsoft's open-source software development kit for building applications powered by large language models, the kind of AI behind chat assistants. Its goal is to let developers add LLM features to their apps and, more recently, to build full "agents", programs that can take instructions, call tools, remember things, and work together with other agents to get a task done. The README opens with an important notice that Semantic Kernel has been succeeded by Microsoft Agent Framework, the enterprise-ready 1.0 release, and links to a migration guide; Semantic Kernel itself is still the project this repository hosts. The SDK is described as model-agnostic, meaning it does not lock you to one AI provider. It has built-in connectors for OpenAI, Azure OpenAI, Hugging Face, and NVIDIA, and can run models locally through Ollama, LMStudio, or ONNX. It also integrates with vector databases such as Azure AI Search, Elasticsearch, and Chroma for giving an agent long-term memory. Key features called out include an agent framework with tools, memory, and planning; multi-agent orchestration; a plugin system that can wrap native code, prompt templates, OpenAPI specifications, or Model Context Protocol (MCP) endpoints; multimodal support for text, vision, and audio; and a process framework for modeling business workflows. It ships for three runtimes: a Python package (semantic-kernel on PyPI, Python 3.10 and up), a .NET NuGet package (Microsoft.SemanticKernel.NET 10.0 and up), and a Java edition (JDK 17 and up). Windows, macOS, and Linux are all supported. The quickstart examples show how to create a ChatCompletionAgent in both Python and C#, then extend it with a plugin class whose methods are marked as kernel functions so the model can call them, with structured output via Pydantic models on the Python side. You would use Semantic Kernel when you want a single, enterprise-oriented SDK for plugging LLMs and agentic behavior into existing Python.NET, or Java applications. It is MIT-licensed.

Copy-paste prompts

Prompt 1
Show me how to create a simple agent in Semantic Kernel that can call a custom plugin to fetch user data.
Prompt 2
How do I set up Semantic Kernel to work with multiple AI providers like OpenAI and Azure OpenAI?
Prompt 3
Walk me through building a multi-agent system where one agent gathers information and another summarizes it.
Prompt 4
How do I connect Semantic Kernel to a vector database for retrieval-augmented generation?
Prompt 5
What's the migration path from Semantic Kernel to Microsoft Agent Framework?
Open on GitHub → Explain another repo

Generated 2026-05-21 · Model: sonnet-4-6 · Verify against the repo before relying on details.