Semantic Kernel<\/strong> is a model-agnostic orchestration framework that lets you connect LLMs, tools, memory, and business logic into AI agents and workflows. You create a kernel, register AI services and plugins (native code and prompt functions), optionally add vector memory, and then invoke functions or agents that can plan and call tools automatically. Semantic Kernel is designed for building enterprise-grade AI applications, from simple chatbots to complex multi-agent systems.<\/p>\n\n\n\n Semantic Kernel (SK)<\/strong> is an open-source AI orchestration framework from Microsoft. It provides a lightweight runtime and SDK for integrating large language models (LLMs) with your code, data, and services.<\/p>\n\n\n\n Key characteristics:<\/p>\n\n\n\n Semantic Kernel is commonly used for:<\/p>\n\n\n\n Semantic Kernel helps you:<\/p>\n\n\n\n It is especially useful when you want a structured, extensible way to build AI applications on top of LLMs.<\/p>\n\n\n\n Semantic Kernel is Microsoft\u2019s open-source, model-agnostic SDK that lets you plug in LLMs, build prompt functions and tools, add vector memory, and orchestrate single- or multi-agent workflows with planning and tool-calling. Learn AI & ML with HCL GUVI\u2019s<\/em> Artificial Intelligence and Machine\u00a0 Learning course<\/em><\/a>.\u00a0<\/em><\/p>\n\n\n\n Before writing code, you need to understand four core ideas.<\/p>\n\n\n\n The kernel<\/strong> is the central runtime in Semantic Kernel.<\/p>\n\n\n\n You can think of the kernel as the \u201capp container\u201d for your AI logic.<\/p>\n\n\n\n Plugins<\/strong> are collections of functions that the kernel can invoke.<\/p>\n\n\n\n There are two main types:<\/p>\n\n\n\n Plugins enable the LLM to:<\/p>\n\n\n\n Prompts<\/strong> are the instructions and context you send to the LLM.<\/p>\n\n\n\n In Semantic Kernel:<\/p>\n\n\n\n This allows you to treat prompts as first-class functions that can be composed, tested, and versioned.<\/p>\n\n\n\n Memory<\/strong> in Semantic Kernel refers to vector-based retrieval over your data.<\/p>\n\n\n\n Memory enables your AI to answer questions grounded in your own documents, not just its pre-trained knowledge.<\/p>\n\n\n\n This section walks you through a minimal end-to-end setup using Python. Equivalent patterns exist in .NET and Java.<\/p>\n\n\n\n Install the Semantic Kernel package with your preferred AI provider.<\/p>\n\n\n\n For OpenAI:<\/p>\n\n\n\n bash<\/p>\n\n\n\n pip install semantic-kernel openai python-dotenv<\/p>\n\n\n\n For Azure OpenAI:<\/p>\n\n\n\n bash<\/p>\n\n\n\n Optionally, install Jupyter support for notebooks:<\/p>\n\n\n\n bash<\/p>\n\n\n\n Create a .env file in your project directory:<\/p>\n\n\n\n text<\/p>\n\n\n\n Never hard-code secrets in your code. Load them from environment variables.<\/p>\n\n\n\n In a Python script or notebook:<\/p>\n\n\n\n python<\/p>\n\n\n\n Now you have a kernel with an LLM <\/a>service registered.<\/p>\n\n\n\n Create a basic prompt that uses the LLM.<\/p>\n\n\n\n python<\/p>\n\n\n\n This defines a Summarize function in a TextPlugin plugin.<\/p>\n\n\n\n Run the function with some input text.<\/p>\n\n\n\n python<\/p>\n\n\n\n The LLM will generate a short summary based on your prompt.<\/p>\n\n\n\n Native plugins let the LLM call your own code<\/a>.<\/p>\n\n\n\n Create a class with methods that perform actions.<\/p>\n\n\n\n python<\/p>\n\n\n\n Each method decorated with @kernel_function becomes callable by the kernel and, if enabled, by the LLM.<\/p>\n\n\n\n python<\/p>\n\n\n\n Now the kernel knows about TimePlugin.get_current_time and TimePlugin.add_numbers.<\/p>\n\n\n\n Create a prompt that can call these functions.<\/p>\n\n\n\n python<\/p>\n\n\n\n With function choice enabled, the LLM can decide to call get_current_time or add_numbers when appropriate.<\/p>\n\n\n\n python<\/p>\n\n\n\n The LLM may call the native functions to answer the questions.<\/p>\n\n\n\n Memory lets the kernel retrieve relevant documents from a vector <\/a>store.<\/p>\n\n\n\n For demonstration, use the built-in volatile memory store. In production, use a real vector database<\/a>.<\/p>\n\n\n\n python<\/p>\n\n\n\n python<\/p>\n\n\n\n These documents are embedded and stored for retrieval.<\/p>\n\n\n\n python<\/p>\n\n\n\n When invoked, the prompt <\/a>will search memory and ground the answer in your documents.<\/p>\n\n\n\n python<\/p>\n\n\n\n The LLM uses retrieved snippets to answer based on your data.<\/p>\n\n\n\n Semantic Kernel supports agent patterns where the LLM can plan, call tools, and collaborate with other agents.<\/p>\n\n\n\n A simple agent is a kernel with:<\/p>\n\n\n\n You already built this with AssistantPlugin and tool-enabled prompts.<\/p>\n\n\n\n For more complex scenarios, Semantic Kernel provides agent orchestration patterns:<\/p>\n\n\n\n Typical steps:<\/p>\n\n\n\n Example conceptual pattern (Python-like pseudocode):<\/p>\n\n\n\n python<\/p>\n\n\n\nTL;DR Summary Box<\/strong><\/h2>\n\n\n\n
\n
What Is Semantic Kernel?<\/strong><\/h2>\n\n\n\n
![\"What](\"https:\/\/www.guvi.in\/blog\/wp-content\/uploads\/2026\/08\/image-566.png\" "\\"\\"")
<\/figure>\n\n\n\n\n
\n
Why Use Semantic Kernel?<\/strong><\/h3>\n\n\n\n
\n
Core Concepts: Kernel, Plugins, Prompts, and Memory<\/strong><\/h2>\n\n\n\n
![\"Core](\"https:\/\/www.guvi.in\/blog\/wp-content\/uploads\/2026\/08\/image-567.png\" "\\"\\"")
<\/figure>\n\n\n\n1. Kernel<\/strong><\/h2>\n\n\n\n
\n
2. Plugins<\/strong><\/h2>\n\n\n\n
\n
\n
3. Prompts and Prompt Functions<\/strong><\/h2>\n\n\n\n
\n
4. Memory<\/strong><\/h2>\n\n\n\n
\n
Step-by-Step: Setting Up Semantic Kernel (Python)<\/strong><\/h2>\n\n\n\n
Step 1: Install the SDK<\/strong><\/h3>\n\n\n\n
pip install semantic-kernel azure-identity python-dotenv<\/code><\/pre>\n\n\n\npip install ipykernel<\/code><\/pre>\n\n\n\nStep 2: Configure Your Environment<\/strong><\/h3>\n\n\n\n
OPENAI_API_KEY=sk-...\n\n# or for Azure OpenAI:\n\n# AZURE_OPENAI_ENDPOINT=https:\/\/your-resource.openai.azure.com\/\n\n# AZURE_OPENAI_DEPLOYMENT_NAME=your-deployment\n\n# AZURE_OPENAI_API_KEY=...<\/code><\/pre>\n\n\n\nStep 3: Create the Kernel<\/strong><\/h3>\n\n\n\n
import asyncio\n\nimport os\n\nfrom dotenv import load_dotenv\n\nimport semantic_kernel as sk\n\nfrom semantic_kernel.connectors.ai.open_ai import OpenAIChatCompletion\n\nload_dotenv()\n\napi_key = os.getenv(\"OPENAI_API_KEY\")\n\nmodel_id = \"gpt-4o-mini\" # or your preferred model\n\nkernel = sk.Kernel()\n\nchat_service = OpenAIChatCompletion(\n\n service_id=\"chat\",\n\n ai_model_id=model_id,\n\n api_key=api_key\n\n)\n\nkernel.add_service(chat_service)<\/code><\/pre>\n\n\n\nStep 4: Define a Simple Prompt Function<\/strong><\/h3>\n\n\n\n
from semantic_kernel.functions import KernelFunctionFromPrompt\n\nsummarize_prompt = \"\"\"\n\nSummarize the following text in 2-3 sentences.\n\nText: {{$input}}\n\nSummary:\n\n\"\"\"\n\nsummarize_function = KernelFunctionFromPrompt(\n\n function_name=\"Summarize\",\n\n plugin_name=\"TextPlugin\",\n\n prompt=summarize_prompt\n\n)\n\nkernel.add_function(summarize_function)<\/code><\/pre>\n\n\n\nStep 5: Invoke the Function<\/strong><\/h3>\n\n\n\n
text = \"\"\"\n\nSemantic Kernel is Microsoft's open-source AI orchestration framework.\n\nIt lets developers integrate LLMs with code, data, and services using plugins,\n\nprompts, and memory. It supports .NET, Python, and Java.\n\n\"\"\"\n\nresult = asyncio.run(\n\n kernel.invoke(\n\n summarize_function,\n\n input=text\n\n )\n\n)\n\nprint(str(result))<\/code><\/pre>\n\n\n\nAdding Native Plugins (Tools)<\/strong><\/h2>\n\n\n\n
Step 6: Define a Native Plugin Class<\/strong><\/h3>\n\n\n\n
text = \"\"\"\n\nSemantic Kernel is Microsoft's open-source AI orchestration framework.\n\nIt lets developers integrate LLMs with code, data, and services using plugins,\n\nprompts, and memory. It supports .NET, Python, and Java.\n\n\"\"\"\n\nresult = asyncio.run(\n\n kernel.invoke(\n\n summarize_function,\n\n input=text\n\n )\n\n)\n\nprint(str(result))<\/code><\/pre>\n\n\n\nStep 7: Register the Plugin<\/strong><\/h3>\n\n\n\n
time_plugin = TimePlugin()\n\nkernel.add_plugin(time_plugin, plugin_name=\"TimePlugin\")<\/code><\/pre>\n\n\n\nStep 8: Use the Plugin in a Prompt<\/strong><\/h3>\n\n\n\n
tool_prompt = \"\"\"\n\nYou are a helpful assistant. You can use the following functions:\n\n- TimePlugin.get_current_time: Get the current date and time.\n\n- TimePlugin.add_numbers: Add two numbers.\n\nUser: {{$input}}\n\nAssistant:\n\n\"\"\"\n\ntool_function = KernelFunctionFromPrompt(\n\n function_name=\"Assistant\",\n\n plugin_name=\"AssistantPlugin\",\n\n prompt=tool_prompt,\n\n prompt_execution_settings=sk.PromptExecutionSettings(\n\n function_choice_behavior=sk.FunctionChoiceBehavior.Auto()\n\n )\n\n)\n\nkernel.add_function(tool_function)<\/code><\/pre>\n\n\n\nStep 9: Invoke the Tool-Enabled Function<\/strong><\/h3>\n\n\n\n
questions = [\n\n \"What time is it now?\",\n\n \"Add 12.5 and 7.3 for me.\"\n\n]\n\nfor q in questions:\n\n result = asyncio.run(\n\n kernel.invoke(\n\n tool_function,\n\n input=q\n\n )\n\n )\n\n print(\"Q:\", q)\n\n print(\"A:\", str(result))\n\n print(\"-\" * 40)<\/code><\/pre>\n\n\n\nAdding Memory for Retrieval-Augmented Generation<\/strong><\/h2>\n\n\n\n
Step 10: Set Up a Vector Store (Example with In-Memory)<\/strong><\/h3>\n\n\n\n
from semantic_kernel.connectors.memory.volatile import VolatileMemoryStore\n\nfrom semantic_kernel.core_plugins import TextMemoryPlugin\n\nmemory_store = VolatileMemoryStore()\n\nmemory_plugin = TextMemoryPlugin(\n\n memory_store=memory_store,\n\n embeddings_service=chat_service # or a dedicated embedding service\n\n)\n\nkernel.add_plugin(memory_plugin, plugin_name=\"Memory\")<\/code><\/pre>\n\n\n\nStep 11: Save Documents to Memory<\/strong><\/h3>\n\n\n\n
docs = [\n\n \"Semantic Kernel is an AI orchestration SDK from Microsoft.\",\n\n \"It supports .NET, Python, and Java for building AI agents.\",\n\n \"Plugins let LLMs call native code and external APIs.\",\n\n]\n\nfor i, doc in enumerate(docs):\n\n asyncio.run(\n\n memory_plugin.save_information(\n\n text=doc,\n\n id=f\"doc_{i}\",\n\n collection=\"sk_docs\"\n\n )\n\n )<\/code><\/pre>\n\n\n\nStep 12: Query Memory in a Prompt<\/strong><\/h3>\n\n\n\n
rag_prompt = \"\"\"\n\nYou are a helpful assistant. Use the Memory plugin to retrieve relevant information.\n\nRelevant information:\n\n{{Memory.search collection='sk_docs' query=$input}}\n\nUser question: {{$input}}\n\nAnswer based on the retrieved information:\n\n\"\"\"\n\nrag_function = KernelFunctionFromPrompt(\n\n function_name=\"RagAssistant\",\n\n plugin_name=\"RagPlugin\",\n\n prompt=rag_prompt\n\n)\n\nkernel.add_function(rag_function)<\/code><\/pre>\n\n\n\nStep 13: Run a RAG Query<\/strong><\/h3>\n\n\n\n
question = \"What is Semantic Kernel and what languages does it support?\"\n\nresult = asyncio.run(\n\n kernel.invoke(\n\n rag_function,\n\n input=question\n\n )\n\n)\n\nprint(\"Q:\", question)\n\nprint(\"A:\", str(result))<\/code><\/pre>\n\n\n\nBuilding AI Agents and Multi-Agent Workflows<\/strong><\/h2>\n\n\n\n
Single AI Agent<\/strong><\/h2>\n\n\n\n
\n
Multi-Agent Orchestration<\/strong><\/h2>\n\n\n\n
\n
\n
# Define agents\n\nresearch_agent = create_agent(\"ResearchAgent\", research_plugins)\n\nwriter_agent = create_agent(\"WriterAgent\", writing_plugins)\n\nreviewer_agent = create_agent(\"ReviewerAgent\", review_plugins)\n\n# Create orchestration\n\norchestration = SequentialOrchestration(\n\n agents=[research_agent, writer_agent, reviewer_agent]\n\n)\n\n# Run\n\nresult = asyncio.run(\n\n orchestration.invoke(\"Write a short article about Semantic Kernel.\")\n\n)<\/code><\/pre>\n\n\n\n