How to beginner · 3 min read

How to create native functions in Semantic Kernel

Q: How to create native functions in Semantic Kernel

In Semantic Kernel, create native functions by defining Python functions and registering them with the Kernel instance using register_native_function. These functions can then be invoked from AI skills or prompts to extend capabilities with custom logic.

Quick answer

In Semantic Kernel, create native functions by defining Python functions and registering them with the Kernel instance using register_native_function. These functions can then be invoked from AI skills or prompts to extend capabilities with custom logic.

PREREQUISITES

Python 3.8+
pip install semantic-kernel
OpenAI API key (free tier works)
pip install openai>=1.0

Setup

Install the semantic-kernel Python package and set your OpenAI API key as an environment variable.

Install Semantic Kernel: pip install semantic-kernel
Set environment variable: export OPENAI_API_KEY='your_api_key' (Linux/macOS) or setx OPENAI_API_KEY "your_api_key" (Windows)

bash

pip install semantic-kernel

Step by step

Define a native Python function and register it with the Kernel. Then invoke it via the kernel's skills interface.

python

import os
import semantic_kernel as sk
from semantic_kernel.connectors.ai.open_ai import OpenAIChatCompletion

# Initialize kernel and add OpenAI chat service
kernel = sk.Kernel()
kernel.add_service(OpenAIChatCompletion(
    service_id="chat",
    api_key=os.environ["OPENAI_API_KEY"],
    ai_model_id="gpt-4o-mini"
))

# Define a native Python function
@sk.native_function("greet")
def greet(name: str) -> str:
    return f"Hello, {name}! Welcome to Semantic Kernel native functions."

# Register the native function with the kernel
kernel.register_native_function(greet)

# Invoke the native function
result = kernel.invoke_native_function("greet", "Alice")
print(result)  # Output: Hello, Alice! Welcome to Semantic Kernel native functions.

output

Hello, Alice! Welcome to Semantic Kernel native functions.

Common variations

You can create async native functions by defining async Python functions and registering them similarly. Also, native functions can accept multiple parameters and return complex types.

Example async native function:

python

import asyncio

@sk.native_function("async_greet")
async def async_greet(name: str) -> str:
    await asyncio.sleep(0.1)  # simulate async work
    return f"Hello asynchronously, {name}!"

kernel.register_native_function(async_greet)

# To invoke async native function use asyncio
import asyncio
result = asyncio.run(kernel.invoke_native_function("async_greet", "Bob"))
print(result)  # Output: Hello asynchronously, Bob!

output

Hello asynchronously, Bob!

Troubleshooting

If you see AttributeError on register_native_function, ensure you are using semantic-kernel version 1.0 or later.
If native functions do not execute, verify your function is decorated with @sk.native_function and properly registered.
For async functions, always invoke with an event loop like asyncio.run().

✅

Key Takeaways

Use @sk.native_function decorator to define native functions in Semantic Kernel.
Register native functions with kernel.register_native_function before invoking.
Async native functions require async def and invocation via asyncio.run().

Verified 2026-04 · gpt-4o-mini

Verify ↗