DEV Community

Alex Spinov
Alex Spinov

Posted on

Semantic Kernel Has a Free API: Build AI Agents with Microsoft's Framework

#ai #microsoft #python #machinelearning

What is Semantic Kernel?

Semantic Kernel is Microsoft's open-source SDK for building AI agents and integrating LLMs into applications. It's the same framework powering Microsoft 365 Copilot, and it supports C#, Python, and Java.

Why Semantic Kernel Over LangChain?

While LangChain dominates the Python ecosystem, Semantic Kernel offers advantages:

  • Enterprise-grade — battle-tested in Microsoft 365 Copilot at massive scale
  • Multi-language — native C#, Python, Java support (not just Python wrappers)
  • Plugin architecture — compose AI capabilities like LEGO blocks
  • Built-in planning — AI automatically chains plugins to achieve goals
  • OpenAI + Azure + local models — swap providers without code changes

Quick Start (Python) 

pip install semantic-kernel
Enter fullscreen mode Exit fullscreen mode 
import semantic_kernel as sk
from semantic_kernel.connectors.ai.open_ai import OpenAIChatCompletion

kernel = sk.Kernel()
kernel.add_service(OpenAIChatCompletion(
    service_id="chat",
    ai_model_id="gpt-4",
    api_key="your-api-key"
))

# Create a semantic function
prompt = """Summarize the following text in 3 bullet points:
{{$input}}"""

summarize = kernel.add_function(
    function_name="summarize",
    plugin_name="TextPlugin",
    prompt=prompt
)

result = await kernel.invoke(summarize, input="Your long text here...")
print(result)
Enter fullscreen mode Exit fullscreen mode

Build a Plugin System 

from semantic_kernel.functions import kernel_function

class WebSearchPlugin:
    @kernel_function(description="Search the web for information")
    def search(self, query: str) -> str:
        # Your search implementation
        return f"Results for: {query}"

class MathPlugin:
    @kernel_function(description="Calculate mathematical expressions")
    def calculate(self, expression: str) -> str:
        return str(eval(expression))  # simplified

# Register plugins
kernel.add_plugin(WebSearchPlugin(), "web")
kernel.add_plugin(MathPlugin(), "math")
Enter fullscreen mode Exit fullscreen mode

Auto-Planning: Let AI Chain Plugins 

from semantic_kernel.planners import SequentialPlanner

planner = SequentialPlanner(kernel)

# AI automatically figures out which plugins to call
plan = await planner.create_plan(
    "Find the current population of Tokyo and calculate what percentage it is of Japan's total population"
)

result = await plan.invoke(kernel)
print(result)  # AI chains web.search → math.calculate
Enter fullscreen mode Exit fullscreen mode

Memory and RAG 

from semantic_kernel.connectors.memory.azure_cognitive_search import AzureCognitiveSearchMemoryStore

memory = AzureCognitiveSearchMemoryStore(
    vector_size=1536,
    search_endpoint="https://your-search.search.windows.net",
    admin_key="your-key"
)

# Save memories
await kernel.memory.save_information(
    collection="docs",
    text="Semantic Kernel supports multiple AI providers",
    id="doc1"
)

# Recall relevant memories
results = await kernel.memory.search("docs", "What AI providers are supported?")
Enter fullscreen mode Exit fullscreen mode

Semantic Kernel vs LangChain vs CrewAI

Feature Semantic Kernel LangChain CrewAI
Languages C#, Python, Java Python, JS Python
Enterprise backing Microsoft LangChain Inc Community
Plugin system Native Tools/Chains Tools
Auto-planning Built-in LangGraph Built-in
Azure integration Native Adapter Manual
Production scale Microsoft 365 Various Early

Real-World Use Case

An enterprise client needed to build an internal knowledge assistant. They tried LangChain first — it worked for prototyping but broke at scale. Switching to Semantic Kernel with Azure OpenAI gave them: native .NET integration with their existing stack, built-in retry/throttling for Azure, and the planner automatically composed 12 plugins into complex workflows.

Building AI agents for your team? I specialize in production AI systems with verified data sources. Contact spinov001@gmail.com or explore my automation tools on Apify.

Top comments (0)