What does this Semantic Kernel MCP connector do?

It provides 8 Python sample scripts demonstrating how to integrate Model Context Protocol (MCP) servers with Semantic Kernel agents, covering scenarios like local agents with local MCP servers, HTTP-based MCP plugins, Azure AI Agents with MCP plugins, and agent-to-agent MCP connections.

When should I use these Semantic Kernel MCP samples?

Use these samples when you want a Semantic Kernel agent (local or Azure AI Agent) to consume tools exposed by an MCP server, whether the server runs locally, over HTTP, or as part of an Azure AI Agent deployment. Choose the sample that matches your specific deployment topology.

How do I install and run these samples?

Each sample is a standalone Python script located in Semantic Kernel's `python/samples/concepts/mcp` directory. Install Semantic Kernel's Python package and any MCP server you want to connect to, then run the script matching your deployment shape (local agent + local server, HTTP MCP plugin, or Azure AI Agent + MCP).

What deployment topologies do these samples cover?

The samples cover local agent with local MCP server, Azure AI Agent with local MCP server, Azure AI Agent with MCP plugin, HTTP-based MCP plugin access, and agent-to-agent MCP connections.

MCP Connector

Integrate Semantic Kernel with MCP Servers

Name: Semantic Kernel - Mcp
Availability: OnlineOnly
Author: Semantic Kernel

8 Semantic Kernel samples for MCP integration: connecting agents to MCP servers as plugins, over HTTP, and with Azure AI Agents.

Connect

Works with azure ai

Semantic Kernel

Own this? Claim it

Maintainer of this project? Claim this page to edit the listing.

Spark score

out of 100

Updated last month

Version python-1.43.1

Models

universal

Add to Favorites

Why it matters

Connect Semantic Kernel applications to Model Context Protocol (MCP) servers, enabling agents and plugins to interact with external data sources and services.

Outcomes

What it gets done

Integrate MCP servers with Semantic Kernel agents.

Develop and deploy MCP plugins for Semantic Kernel.

Sample data from MCP servers for agent use.

Connect to local and Azure AI MCP servers.

Install

Add it to your toolbox

Run in your project directory:

curl -fsSL https://spark.entire.vc/get/sk-concept-mcp | bash

Overview

Semantic Kernel - Mcp

Eight Semantic Kernel Python samples for MCP integration, covering local and HTTP MCP servers, MCP sampling, and Azure AI Agent connections. Use to connect a Semantic Kernel or Azure AI Agent to MCP servers; pick the sample matching your specific local/HTTP/Azure deployment shape.

What it does

This is a set of 8 Semantic Kernel Python concept samples demonstrating Model Context Protocol (MCP) integration: an agent using an MCP plugin over HTTP (agent_with_http_mcp_plugin.py), an agent acting as/using an MCP agent (agent_with_mcp_agent.py), an agent using an MCP plugin (agent_with_mcp_plugin.py), an agent using MCP sampling (agent_with_mcp_sampling.py), an Azure AI Agent connected to a local MCP server (azure_ai_agent_with_local_server.py), an Azure AI Agent using an MCP plugin (azure_ai_agent_with_mcp_plugin.py), a local agent connected to a local MCP server (local_agent_with_local_server.py), and using MCP itself as a plugin (mcp_as_plugin.py).

When to use - and when NOT to

Use these when you want a Semantic Kernel agent (local or Azure AI Agent) to consume tools exposed by an MCP server - whether that server runs locally, over HTTP, or as part of an Azure AI Agent deployment. The variety here lets you match the sample to your exact deployment shape: local agent + local server, Azure AI Agent + MCP plugin, or HTTP-based MCP access.

Not needed if you're building custom tool integrations outside the MCP standard - these samples specifically cover the MCP protocol path for connecting agents to external tool servers, not a general-purpose tool-calling tutorial.

Capabilities

Connecting a Semantic Kernel agent to an MCP server as a plugin (local process or over HTTP), using MCP sampling, agent-to-agent MCP connections, and connecting an Azure AI Agent to either a local MCP server or an MCP plugin.

How to install

Each sample is a standalone Python script in Semantic Kernel's python/samples/concepts/mcp directory; run the one matching your deployment shape (local agent + local server, HTTP MCP plugin, or Azure AI Agent + MCP) after installing Semantic Kernel's Python package and any MCP server you intend to connect to.

Who it's for

Developers connecting Semantic Kernel agents - local or Azure AI Agent-based - to MCP servers, who need a sample matching their specific deployment topology (local, HTTP, or Azure).

Source README

Model Context Protocol (MCP) integration: MCP servers, agents, plugins, and sampling

Examples (8 files):

agent_with_http_mcp_plugin.py
agent_with_mcp_agent.py
agent_with_mcp_plugin.py
agent_with_mcp_sampling.py
azure_ai_agent_with_local_server.py
azure_ai_agent_with_mcp_plugin.py
local_agent_with_local_server.py
mcp_as_plugin.py

FAQ

Common questions

Discussion