Semantic Kernel: Microsoft's AI SDK, now handing over to Agent Framework
Semantic Kernel is Microsoft's open source SDK (MIT license) for bringing AI models and agents into C#, Python or Java applications. Microsoft now treats it as superseded by Microsoft Agent Framework: it is still maintained, but most new work goes to its successor.
Data as of October 22, 2026
Merged into another product: it no longer evolves under this name.
Key figures
| Developer | Microsoft |
|---|---|
| Type | Code |
| Languages | C#, Python, Java |
| Open source | Yes |
| License | MIT |
| First release | March 2023 |
| Suggested level | Intermediate |
| Use cases | chatbot, multi-agent, RAG |
| Self-hostable | Yes |
| Business model | Free |
| Cloud offering | Not disclosed |
| Multi-agent | Yes |
| MCP support | Yes |
| Bring your own model | Yes |
| Official site | Official site · Documentation · Repository |
GitHub
| Repository | microsoft/semantic-kernel |
|---|---|
| GitHub stars | 28,688 |
| Forks | 4,832 |
| Latest code activity | October 22, 2026 |
| Latest release | python-1.45.0 (October 6, 2026) |
GitHub activity checked on October 22, 2026.
Where Semantic Kernel comes from
Microsoft launched Semantic Kernel in 2023 to help developers of regular applications, especially in C#, plug an AI model into their code without reinventing everything. The idea: a “kernel” that orchestrates plugins, meaning functions from your app, prompt templates or APIs described in OpenAPI, which the model can call through tool calling. The SDK later added AI agents, multi-agent orchestration, MCP support, vector database connectors and a Process Framework for modeling business processes.
Status as of October 2026
The official README puts it plainly: “Semantic Kernel is now Microsoft Agent Framework.” The successor, built by the same team, carries over Semantic Kernel’s enterprise features and the multi-agent abstractions from AutoGen.
According to the official blog post from October 2025:
- version 1.x remains maintained for critical bugs and security issues;
- that maintenance is promised for at least one year after Agent Framework’s general availability (version 1.0 shipped on April 2, 2026);
- most new features are being built for Agent Framework.
At the time of writing, new Semantic Kernel releases are still coming out for both Python and .NET.
Starting fresh or already using it
Starting from scratch: follow Microsoft’s advice and begin with Agent Framework, which offers explicit workflows and a long-term support commitment.
You already have a Semantic Kernel app: it keeps working and keeps receiving important fixes. Use this window to read the official migration guide, list the plugins and agents you will need to port, and schedule the move before the maintenance commitment ends.
Strengths and limits
Strengths: mature .NET support, a clear plugin-based approach, many model and vector database connectors, and the option to run with local models.
Limits: this is a project at the end of its cycle, so investing heavily in it today means a migration later. New capabilities in the field, such as graph-based workflows or the harness agent, are landing in Agent Framework rather than here. If you are not tied to Microsoft, LangChain is another widely used option for connecting models, tools and data.
Frameworks of the same type
- Agno · multi-agent, workflows, RAG
- Mastra · multi-agent, workflows, RAG
- Amazon Bedrock Agents / AgentCore · multi-agent, RAG, workflows
- BeeAI Framework · multi-agent, workflows, RAG
- CAMEL · multi-agent, data, automation
- Google Agent Development Kit (ADK) · multi-agent, workflows, chatbot
Frequently asked questions
Has Semantic Kernel been abandoned?
Not as of today. Microsoft committed to fixing critical bugs and security issues in version 1.x for at least one year after Agent Framework became generally available in April 2026. New releases still ship, but most new features are reserved for Agent Framework.
Should I migrate to Microsoft Agent Framework?
For a new project, that is Microsoft's recommendation. For an existing app there is no overnight rush, but an official migration guide exists and it is wise to plan the move while maintenance is still guaranteed.
Does Semantic Kernel work outside Azure?
Yes. According to the official repository, it connects to OpenAI, Azure OpenAI, Hugging Face or NVIDIA, and can run with local models through Ollama, LM Studio or ONNX.
Which languages are supported?
C# (.NET) and Python in the main repository, plus Java in a separate repository. The project has historically been especially popular with .NET developers.
Data sources
- github.com
- github.com
- devblogs.microsoft.com
- learn.microsoft.com
- learn.microsoft.com
- GitHub — Repository metadata: GitHub
Data as of October 22, 2026
Report an error
Does a figure, a date or a fact look wrong? Tell us: every report is checked.