Semantic Kernel

Every company has a mission. What's ours? To empower every person and every organization to achieve more. We believe technology can and should be a force for good and that meaningful innovation contributes to a brighter world in the future and today. Our culture doesn’t just encourage curiosity; it embraces it. Each day we make progress together by showing up as our authentic selves. We show up with a learn-it-all mentality. We show up cheering on others, knowing their success doesn't diminish our own. We show up every day open to learning our own biases, changing our behavior, and inviting in differences. Because impact matters. Microsoft operates in 190 countries and is made up of approximately 228,000 passionate employees worldwide.

Agent Framework

Visit Website

Founded

2025

Location

Redmond, Washington

Employees

225889

Funding

OSS

Semantic Kernel (Microsoft) — Open-Source SDK for AI Agents

Overview

**Semantic Kernel (SK)** is Microsoft’s open-source SDK for building AI agents and agentic applications in **.NET (C#), Python, and Java**. It’s model-agnostic and provides a typed, testable way to connect LLMs to tools, plugins, memory, and enterprise data—well-aligned with Microsoft’s ecosystem and enterprise development patterns.

Official docs on [Microsoft Learn](https://learn.microsoft.com/en-us/semantic-kernel/overview/)

Source, issues, and releases on [GitHub](https://github.com/microsoft/semantic-kernel)

Agent Framework docs within SK: [Agent Framework](https://learn.microsoft.com/en-us/semantic-kernel/frameworks/agent/)

In 2025, Microsoft introduced the **Microsoft Agent Framework (MAF)**—a unified SDK/runtime combining learnings from SK and AutoGen. Microsoft provides **migration guidance** from SK’s Agent Framework to MAF. Teams starting new builds should evaluate both:

MAF migration guide: [From Semantic Kernel](https://learn.microsoft.com/en-us/agent-framework/migration-guide/from-semantic-kernel/)

Key Capabilities

**Agent orchestration:** planning, tool use, multi-step workflows, and event-driven coordination

**RAG & memory:** connectors to popular vector stores; chunking, embeddings, and retrieval patterns

**Tooling & function calling:** clean, typed interfaces for tool execution and function calls

**Streaming chat & multi-agent patterns:** stateful sessions, history, and context management

**Observability:** OpenTelemetry support for tracing and metrics

**Enterprise fit:** typed SDKs, testability, and patterns familiar to Microsoft stack teams

Tech Stack & Integrations

Languages: **.NET (C#), Python, Java**

License: **MIT**

Models (model-agnostic): **Azure OpenAI, OpenAI, GitHub Models, Google, Mistral, Hugging Face, Ollama, ONNX**

Docs: [Chat completion](https://learn.microsoft.com/en-us/semantic-kernel/concepts/ai-services/chat-completion/), [Embedding generation](https://learn.microsoft.com/en-us/semantic-kernel/concepts/ai-services/embedding-generation/)

Vector stores: **Azure AI Search, Cosmos DB, Elasticsearch, MongoDB Atlas Vector Search, Pinecone, Postgres/pgvector, Qdrant, others**

Connectors: [Out-of-the-box](https://learn.microsoft.com/en-us/semantic-kernel/concepts/vector-store-connectors/out-of-the-box-connectors/), [Memory stores](https://learn.microsoft.com/en-us/semantic-kernel/concepts/vector-store-connectors/memory-stores)

Cloud & ops: **Azure AI Studio, Azure Functions, Kubernetes** patterns shown in samples

Samples: [Semantic Kernel GitHub](https://github.com/microsoft/semantic-kernel), [Workshop](https://github.com/Azure-Samples/semantic-kernel-workshop)

GitHub stars: ~26.5k (see [Releases page](https://github.com/microsoft/semantic-kernel/releases) for current)

Adoption & Community Sentiment

User feedback highlights SK’s strengths in orchestration and enterprise alignment, with caveats during the transition to MAF.

Pros

**Strong for multi-agent orchestration** when context, memory, and tool boundaries are well-designed — [Reddit discussion](https://www.reddit.com/r/dotnet/comments/1od7lcw/anyone_tried_semantic_kernel_here/)

**Great fit for .NET teams and Microsoft stack**; easy pairing with Azure OpenAI and GitHub Models — [Community feedback](https://www.reddit.com/r/LocalLLaMA/comments/1ghaz6q/just_tried_out_semantic_kernel_in_net/)

**Works locally with Ollama**; straightforward to build local agents in C# — [Developer thread](https://www.reddit.com/r/dotnet/comments/1j10pr5/building_local_ai_agents_with_semantic_kernel_and/)

Cons

**Pace of updates** concerns during the shift toward MAF — [Comparative thread](https://www.reddit.com/r/AI_Agents/comments/1l85499/which_agentic_ai_framework_is_the_best_ms/)

**Learning curve** around multi-user state, history, and session management — [Discussion](https://www.reddit.com/r/dotnet/comments/1fol2jg/semantic_kernel_multiple_users_initial_context/)

**Complexity vs. lighter alternatives** in certain scenarios — [Opinion thread](https://www.reddit.com/r/LocalLLaMA/comments/14cagpc/the_best_framework_currently_going_forward/)

Who It’s For

Enterprise and product teams building **agentic features** in .NET, Python, or Java

Developers invested in **Azure OpenAI, GitHub Models, and Microsoft cloud services**

Teams needing **typed SDKs, observability, and maintainable orchestration** over ad-hoc scripts

Primary Use Cases

AI copilots for internal tools or customer-facing apps

**RAG assistants** over enterprise content using Azure AI Search or other vector stores

**Multi-step agents** that call APIs, databases, and workflows with planning and tool use

**Local agents** for offline/privacy-sensitive work with Ollama

Domain-specific chat and workflow automation inside **.NET line-of-business apps**

Code review assistants, documentation helpers, and support agents

Stories and updates: [Microsoft SK Blog](https://devblogs.microsoft.com/semantic-kernel/)

2025 Positioning: SK vs. Microsoft Agent Framework (MAF)

Microsoft launched **MAF** to unify agent development across SK and AutoGen.

Announcement/migration: [MAF Migration Guide](https://learn.microsoft.com/en-us/agent-framework/migration-guide/from-semantic-kernel/)

Commentary: [Community newsletter](https://newsletter.victordibia.com/p/microsoft-agent-framework-semantic), [Industry post](https://www.linkedin.com/posts/aboutasha_today-the-microsoft-agent-framework-brings-activity-7379155063946223617-Vhh8)

Guidance:

If you’re starting fresh with heavy agent needs, evaluate **MAF** alongside SK’s Agent Framework.

If you’re already on SK, review Microsoft’s **migration guidance** and roadmap alignment.

Pricing & Licensing

**Free and open source** under the **MIT License** — [GitHub repo](https://github.com/microsoft/semantic-kernel)

No trial required. You only pay for chosen external providers (e.g., **Azure OpenAI**).

Learning & Resources

Docs home: [Semantic Kernel Docs](https://learn.microsoft.com/en-us/semantic-kernel/)

Overview: [What is Semantic Kernel?](https://learn.microsoft.com/en-us/semantic-kernel/overview/)

Agent Framework: [SK Agent Framework](https://learn.microsoft.com/en-us/semantic-kernel/frameworks/agent/)

Vector connectors: [Connectors](https://learn.microsoft.com/en-us/semantic-kernel/concepts/vector-store-connectors/out-of-the-box-connectors/)

Migration to MAF: [From SK to MAF](https://learn.microsoft.com/en-us/agent-framework/migration-guide/from-semantic-kernel/)

Samples & workshops: [GitHub repo](https://github.com/microsoft/semantic-kernel), [SK Workshop](https://github.com/Azure-Samples/semantic-kernel-workshop)

Releases and star count: [GitHub Releases](https://github.com/microsoft/semantic-kernel/releases)

Quick Start Checklist

Choose your language SDK: **.NET, Python, or Java**

Select a model provider (e.g., **Azure OpenAI** or **OpenAI**) and set credentials

Pick a vector store (e.g., **Azure AI Search** or **Postgres/pgvector**)

Start with SK samples for **RAG**, **tool calling**, and **multi-agent orchestration**

Add **OpenTelemetry** for observability; plan for **state/session management**

If building long-lived agent systems, review **MAF** and the **migration path** early

If helpful, we can tailor SK/MAF recommendations to your stack, list exact connectors and SDK packages, and sketch a reference architecture for your use case.

Related Companies

AgentGPT

AgentGPT is an open‑source, browser‑run tool from Reworkd that lets you name an agent, set a goal, and have it autonomously plan, research, and execute tasks in iterative loops. It’s designed for fast experimentation and demos, with ready‑made templates for research, branding, and trip planning—no engineering setup required.

BabyAGI

BabyAGI is pioneering the future of autonomous AI through an experimental framework designed for self-building agents. Born from the insight that the most effective path to general autonomous agents is radical simplicity, BabyAGI focuses on creating the minimal viable system capable of building and evolving itself. At its core is functionz, an innovative function framework that revolutionizes how autonomous agents manage their capabilities. This database-driven system stores, manages, and executes functions through an intelligent graph-based architecture that tracks imports, dependencies, and authentication—all with automatic loading and comprehensive logging. BabyAGI provides developers with an intuitive dashboard for seamless function management, real-time updates, and detailed log analysis, making autonomous agent development accessible and transparent. By embracing a self-building philosophy, BabyAGI represents a fundamentally new approach to creating AI systems that can adapt, grow, and improve autonomously.

CrewAI

CrewAI is at the forefront of Agentic AI with its open source, multi-agent framework and cloud platform for building, managing and scaling agentic workflows across the entire organization.

Flowise AI

Flowise is an open source drag & drop tool to build your customized LLM flow. We provide a visual interface to let you build backends for LLM apps used for QnA, summarization and analysis on your documents.

LangChain

LangChain provides the agent engineering platform and open source frameworks developers need to ship reliable agents fast.

LlamaIndex

LlamaIndex empowers developers to build agents that extract insights and take action on complex enterprise documents. It combines industry-leading document parsing and extraction with a trusted framework for building intelligent agents that reason over documents, adapt to business logic, and scale to production. LlamaIndex is loved by developers and trusted by enterprises. Its open source framework is downloaded more than 4M+ every month and has processed more than 200 million documents on LlamaCloud.