Microsoft Copilot Studio ❤️ MCP

What is Microsoft Copilot Studio ❤️ MCP?

This lab provides an MCP server that delivers jokes and demonstrates how to deploy and integrate a Model Context Protocol server with Microsoft Copilot Studio. It is for makers and developers who want to connect AI models to data sources via the MCP standard.

How to use Microsoft Copilot Studio ❤️ MCP?

Clone the template repository, install dependencies with npm install, and run the server locally with npm run build && npm run start, or deploy to Azure using azd up. Then import the MCP connector into Copilot Studio from GitHub, create an agent, and add the server as a tool.

Key features of Microsoft Copilot Studio ❤️ MCP?

• Deploy MCP server locally or to Azure.
• Integrate with Copilot Studio via custom connectors.
• Supports enterprise security like Virtual Network and DLP.
• Provides real-time data access for AI agents.
• Uses the Model Context Protocol standard.
• Includes an example Jokes MCP server.

Use cases of Microsoft Copilot Studio ❤️ MCP?

• Build a joke-telling agent in Copilot Studio.
• Connect external APIs as tools for AI-powered agents.
• Prototype MCP server integration before production.
• Use the same server in Visual Studio Code / GitHub Copilot.

FAQ from Microsoft Copilot Studio ❤️ MCP?

How does MCP differ from connectors?

MCP servers are made available to Copilot Studio using connector infrastructure, so they are better together. MCP leverages the same enterprise security and governance controls as connectors.

What are the prerequisites to run this lab?

Visual Studio Code, Node v22, Git, Azure Developer CLI, an Azure subscription, a GitHub account, a Copilot Studio trial or developer account, and a Power Platform environment with the “Get new features early” toggle enabled.

Where does the data live?

The MCP server runs locally via a dev tunnel or is deployed to Azure Container Apps. Data is accessed in real-time; no permanent storage is described.

What transport and authentication does the server use?

The server uses Streamable HTTP with an /mcp endpoint. Authentication is handled through the connector infrastructure; the lab uses user authentication via the action wizard.

Are there known issues or limitations?

Yes, there are known issues and planned improvements listed at aka.ms/mcsmcp. For example, after importing the connector, a warning and error may appear but can be ignored.