Semantic Kernel
An open-source AI agent SDK for C#, Python and Java. Build workflows with local models through Ollama, LMStudio or ONNX, or connect to cloud providers.
Semantic Kernel is an MIT-licensed SDK for developers adding AI agents to their applications. It supports local models through Ollama, LMStudio and ONNX, alongside cloud services such as OpenAI and Azure OpenAI. You choose the model backend. The SDK runs on Windows, macOS and Linux and supports C#, Python and Java.
Agents can use tools, memory and planning capabilities, while multi-agent workflows let specialist agents collaborate on a task. The Process Framework supports structured business workflows, so the scope extends beyond a chat interface to applications that coordinate several steps or agents.
Plugins connect agents to native code functions, prompt templates, OpenAPI services and Model Context Protocol (MCP) tools. For applications that retrieve stored information, Semantic Kernel integrates with Azure AI Search, Elasticsearch and Chroma. It also supports text, vision and audio inputs.
The choice of backend matters for where processing happens: local deployment uses models on your own hardware, while cloud provider connections send model requests to those services. Observability, security components and filters give developers controls for integrating agents into an application. Microsoft identifies Microsoft Agent Framework as Semantic Kernel's successor and provides a migration guide.
