Skip to content
Agent Guides

Semantic Kernel (Python)

Microsoft’s enterprise SDK for integrating LLMs with plugins, planners, and vector stores. Deep Azure AI integration, MCP client/server, A2A protocol, and bridges into the newer Microsoft Agent Framework.

Install

Terminal window
pip install --pre semantic-kernel

v1.43.0 requires --pre (pip) or --prerelease=allow (uv) — see What’s new.

Version

v1.43.0 · June 2026 · Python 3.10+

Best for

Azure-native enterprise agents, plugin-based architectures, Microsoft 365 integrations.

Zero → Hero

Section titled “Zero → Hero”
1 · Installation & setup Install, auth, first kernel · 15 min
2 · Kernel initialization Kernel builder, services · 15 min
3 · Service configuration OpenAI, Azure OpenAI, HF · 15 min
4 · Semantic functions Prompt-based functions · 20 min
5 · Native functions Python-based functions · 15 min
6 · Plugin development Custom plugins · 20 min
7 · Simple agents ChatCompletionAgent · 15 min
8 · Memory systems Short-term, long-term · 15 min
9 · Vector store v1.34 New vector store API · 20 min
10 · Planners Sequential, Stepwise, Action · 25 min
11 · MCP integration Client + server (2025) · 25 min
12 · A2A protocol Google A2A integration · 20 min
13 · Agent Framework bridge SK ↔ Agent Framework · 15 min
14 · Advanced multi-agent Complex topologies · 30 min
15 · Middleware Hooks, filters · 20 min
16 · Production Deployment, scaling · 30 min

Jump to a topic

Section titled “Jump to a topic”
Kernel initialization
Semantic functions
Plugin development
ChatCompletionAgent
Vector store v1.34
Sequential planner
Stepwise planner
MCP client
MCP server
A2A integration
Middleware
Streaming server

Reference

Section titled “Reference”
Comprehensive (Python) Single-page technical reference.
Advanced multi-agent
Middleware
Recipes
Diagrams
Production guide
Streaming server (FastAPI)

What’s new in v1.43.0 (June 2026)

Section titled “What’s new in v1.43.0 (June 2026)”
  • v1.43.0 (June 2026) — Minor release. semantic-kernel==1.43.0 confirmed latest stable on PyPI (2026-06-04). Install with pip install --pre semantic-kernel or uv pip install --prerelease=allow semantic-kernel as the azure-ai-agents pre-release dependency is expected to persist. Full symbol verification deferred (PyMeta3 build failure blocked install); non-Azure core symbols (Kernel, ChatCompletionAgent, AgentGroupChat, ChatHistory, kernel_function, SelectionStrategy, TerminationStrategy) are assumed stable but unverified — carry-forward for next run.
  • v1.42.0 (May 14) — Minor release. Requires mcp>=1.26.0 (raised from 1.8) and azure-ai-agents>=1.2.0b3 (pre-release dependency; plain pip install semantic-kernel will fail — use pip install --pre semantic-kernel or uv pip install --prerelease=allow semantic-kernel). Non-Azure core symbols (Kernel, ChatCompletionAgent, AgentGroupChat, ChatHistory, kernel_function, SelectionStrategy, TerminationStrategy) verified against installed semantic-kernel==1.42.0 (.routine-envs/check-0514-sk). Known issue: importing semantic_kernel.connectors.ai.open_ai (which includes OpenAIChatCompletion) emits PydanticDeprecatedSince211 from AzureAISearchDataSource at azure_chat_prompt_execution_settings.py:134 — an internal SK issue; the warning does not affect non-Azure connectors at runtime. See risk register in the 2026-05-14 run report.
  • v1.41.3 (April 28) — Patch release; stability improvements.
  • Vector store v1.34 — new API for embedding and retrieval.
  • MCP client + server — full Model Context Protocol support.
  • Google A2A protocol integration.
  • Microsoft Agent Framework bridge for gradual migration.

Install, minimal example

Section titled “Install, minimal example”
import asyncio
from semantic_kernel import Kernel
from semantic_kernel.connectors.ai.open_ai import OpenAIChatCompletion
from semantic_kernel.agents import ChatCompletionAgent


async def main():
    kernel = Kernel()
    kernel.add_service(OpenAIChatCompletion(service_id="chat", ai_model_id="gpt-4o"))


    agent = ChatCompletionAgent(
        service_id="chat",
        kernel=kernel,
        name="Assistant",
        instructions="You are a helpful assistant.",
    )


    async for message in agent.invoke("What is 15 + 27?"):
        print(message.content)


asyncio.run(main())

Ready? Start Zero → Hero Step 1 →

Revision history

Section titled “Revision history”
DateFramework version (installed)Summary of changesReviewer
2026-06-04semantic-kernel 1.43.0Version bumped 1.42.0 → 1.43.0 (minor release); version card updated to June 2026; “What’s new” updated; install card anchor corrected; comprehensive guide header updated; revision history entry added. semantic-kernel==1.43.0 confirmed latest stable on PyPI (2026-06-04); install with --pre flag (or --prerelease=allow in uv) expected to remain necessary due to azure-ai-agents pre-release transitive dep; full symbol verification blocked by env constraint (see risk register).Claude routine
2026-05-14semantic-kernel 1.42.0Version bumped 1.41.3 → 1.42.0; version card updated to May 2026; “What’s new” updated with mcp>=1.26.0 requirement and PydanticDeprecatedSince211 warning note; comprehensive guide header updated; revision history entry added. Non-Azure core symbols verified against installed semantic-kernel==1.42.0 (.routine-envs/check-0514-sk).Claude routine
2026-04-28semantic-kernel 1.41.3Version bumped 1.41.2 → 1.41.3 (patch release); version card and “What’s new” heading updated. PyPI confirms 1.41.3 as latest stable.Claude routine
2026-04-22semantic-kernel 1.41.2Version card corrected from stale v1.36.0 to v1.41.2 (confirmed against PyPI); frameworks.ts updated from Python 1.36.0 to Python 1.41.2.Claude routine
2026-04-21semantic-kernel 1.41.2Index redesigned with Zero → Hero + Jump-to-topic grid.—
April 2026semantic-kernel 1.41.2Vector store v1.34; MCP; A2A; Agent Framework bridge.—

Other languages: .NET.