semantic kernel

It provides a structured framework for building AI agents and copilots, offering abstractions for prompt management, memory and vector search, tool calling (plugins), and multi-agent orchestration that work consistently across different LLM providers.

The SDK's plugin architecture allows developers to expose existing business logic as callable functions that LLMs can invoke autonomously querying a database, calling a REST API, retrieving from a knowledge base without restructuring the underlying codebase.

Semantic Kernel handles the translation between natural language intent and function calls, including automatic function selection from a registered plugin catalog, argument extraction from user messages, and result formatting for the next reasoning step.

Semantic Kernel is the foundation for Microsoft Copilot experiences across Office, Azure, and enterprise products, giving it strong production validation at scale.

It integrates with Azure AI Search, Azure Cosmos DB, and Azure OpenAI Service for organizations on the Microsoft cloud stack, while remaining cloud-agnostic for teams using other providers.

The SDK is open-source under the MIT license, with active development on GitHub and commercial support available through Microsoft's enterprise channels.