Semantic Kernel

Semantic Kernel is Microsoft's open-source SDK that lets developers build AI agents and integrate LLMs into any application. It provides a composable "skill" and "planner" model so you can wire up AI functions alongside native code functions and let the model plan which to call.

Key features

• Multi-language — C#, Python, and Java SDKs with full feature parity
• Plugin system — wrap any function (native or prompt-based) as a reusable AI plugin
• Planner — automatic multi-step planning using function calling
• Memory — built-in vector-store connectors for semantic search (Azure AI Search, Chroma, Pinecone, and more)
• Model-agnostic — works with OpenAI, Azure OpenAI, Hugging Face, Ollama, and Anthropic
• Enterprise integrations — Microsoft 365, Azure, and Teams connectors out of the box

Quick start (Python)

npx ai-supply add semantic-kernel-agent-sdk

# Or install directly
pip install semantic-kernel

import asyncio
from semantic_kernel import Kernel
from semantic_kernel.connectors.ai.open_ai import OpenAIChatCompletion

kernel = Kernel()
kernel.add_service(OpenAIChatCompletion(model_id="gpt-4o", api_key="..."))

summarize = kernel.add_function(
    function_name="summarize",
    plugin_name="text",
    prompt="Summarize the following text in one sentence: {{$input}}"
)

async def main():
    result = await kernel.invoke(summarize, input="Semantic Kernel is great.")
    print(result)

asyncio.run(main())

Curated mirror of the open-source Semantic Kernel project (MIT). Install upstream from the repository.