WonderLaura @iwmentor – Laura Rogers

You’ve probably started hearing the term “MCP” everywhere lately. So what is it, and why should a business user like you care? That’s exactly what we covered on this week’s Power Hour post and video, and I want to break it down here too, because MCP in Copilot Studio is genuinely one of the most exciting no-code capabilities I’ve seen. It’s not just a Microsoft thing, either.

Here’s the thing: you don’t need to be a developer to use it. You just need to understand what it does and how to connect it.

🤔 What Is MCP in Copilot Studio, Anyway?

MCP stands for Model Context Protocol. It’s an open standard that Anthropic introduced in 2024, and it’s now been adopted across the AI world, including by Microsoft. Think of it as USB-C for AI. Just like that one cable can plug into almost anything, MCP gives an AI agent a universal way to connect to your data, your tools, and your workflows.

When you use MCP in Copilot Studio, three pieces work together:

• The MCP host – this is the AI app you’re chatting in, like Copilot Studio or Claude.
• The MCP server – this is what exposes a system’s tools and data in that standard MCP format. Developers build these, but you and I just tap into them.
• Your system – this is SharePoint, Dataverse, ShareGate, or whatever tool actually holds your data and takes the action.

So instead of clicking through five different screens to find a report, you just ask your agent a question, and it goes and gets the answer for you.

⚖️ Flows vs. Agents: Why It Matters for MCP in Copilot Studio

Before you dive into MCP in Copilot Studio, it helps to understand how an agent is different from a Power Automate flow. Because these two concepts get confused a lot, and they solve different problems.

A flow is structured. You design the exact steps, and every time it runs, it follows that same path. It’s predictable, and that’s exactly why you’d choose it for anything that needs guaranteed, repeatable execution.

An agent, on the other hand, is adaptive. It decides which tools to call based on what you ask, and it can self-discover new ones as they become available. As a result, agents are best for fuzzy, multi-step tasks that require judgment rather than a fixed process.

Here’s a simple way to decide which one you need:

• Reach for a flow when the steps never change and timing matters.
• Reach for an agent when the task needs judgment, or when you want natural language help on demand.

The best part? An agent can actually call a flow as one of its own tools. So you get the reasoning of an agent layered on top of the reliability of a flow. That combination is where MCP in Copilot Studio really starts to shine.

🧰 How to Get Started with MCP in Copilot Studio

Adding your first MCP tool takes fewer steps than you’d think. Here’s the process I used to connect ShareGate. Remember a couple of blog posts and power hours ago, when I showed you ShareGate Protect and how you can use it for governance in your tenant? Well, you can chat with it in natural language with the MCP!

1. Create a blank agent in Copilot Studio. You don’t have to configure anything else on the main screen yet. I named my agent ShareGate MCP.

2. Give it a name and description. This helps your team understand what the agent is for once you publish it.

3. Go to the Tools tab and click “Add a tool.”

4. In the Create new section, Choose Model Context Protocol. If you peruse the list below, you’ll see built-in options for SharePoint, Teams, Outlook, Power Apps, and Dataverse, plus third-party ones like DocuSign, and even Zapier.

5. Enter the server URL and authenticate. For example, ShareGate requires OAuth with dynamic discovery, and you’ll need an active ShareGate account to connect. Fill in the server URL, which is https://mcp.sharegate.com/mcp for North America and https://mcp.eu.sharegate.com/mcp for Europe. Scroll down and click CREATE.

Here is the Tools tab after it’s been added:

6. Test it, then publish your agent to a channel like Teams, SharePoint, or Microsoft 365.

One important note: everyone who uses your agent still needs their own credentials for that underlying system. So if a coworker tries your ShareGate agent, they’ll be prompted to log in with their own ShareGate account first. That means MCP in Copilot Studio doesn’t accidentally open up access to information someone shouldn’t see.

You’ll also notice in Copilot Studio that “Workflows” are front and center on the home screen. This graphic shows the differences between Workflows in Copilot Studio and Power Automate flows:

🚀 Real-World Examples of MCP in Copilot Studio

I tested this myself with the ShareGate MCP server, and honestly, it blew me away. I simply asked my agent, “Which are the three largest sites in my tenant?” and it queried ShareGate directly and gave me the answer in seconds. No logging in, no clicking through reports.

I also connected to the ShareGate MCP in my Claude account, and asked it to look at my ShareGate environment and build a PowerPoint for an executive governance meeting, highlighting risks and action items. It generated the entire slide deck from that one prompt. Imagine building that report every month for a governance committee without touching a single dashboard yourself. Also, you could create a skill or template, so that it uses your brand colors and has the same structure each time!! Think of all the time you’ll save, not having to click through the admin screens in Microsoft 365!

The SharePoint MCP works the same way. I asked my agent to list every document library in a specific site, and it found the site, then pulled the library list automatically, narrating each step as it worked. That’s a different experience than chatting with Copilot inside SharePoint itself, because the SharePoint MCP is built around structured actions like create, update, find, and delete. It handles specific, repeatable requests really well, like “create a project issues list with three starter items.”

Speaking of the SharePoint Work IQ MCP, I asked Copilot what the difference is between this and chatting with Copilot about SharePoint data. Here’s a fascinating infographic it created:

🎨 A Few Fun MCP Servers Worth Exploring

Not every MCP server is strictly business. Once you start looking, you’ll find a huge and growing catalog. A few worth knowing about:

• Kit (my newsletter platform) – I can ask about my Power Hour voting results without ever opening the Kit website.
• Spotify – describe a mood, and it builds you a playlist.
• Blender – design a 3D model just by describing it in a prompt.
• Google Maps and Airbnb – community-built servers for travel planning.

The point is this: whatever tools and services you already use, keep an eye out for MCP support. Companies are rolling these out constantly, and often they’ll just show up in your product’s release notes.

💡 Why This Matters for Business Users

MCP in Copilot Studio is a big deal for citizen developers: no-code, business-side power users who don’t have a developer on speed dial. You don’t need anyone to write custom integrations for you anymore. You just add the tool, authenticate, and start chatting.

That’s less time clicking through apps hunting for a report, and more time simply describing the outcome you want. And because MCP lives inside your existing Microsoft 365 tenant, you’re not introducing a brand-new system that has to go through a security review. That’s a fast lane worth taking advantage of.

If you want to go deeper, I cover Copilot Studio for business users inside our Ultimate membership at IW Mentor.

Have you connected an MCP tool inside Copilot Studio yet? I’d love to hear what you’re building. Drop a comment and let me know!

Here’s the associated step-by-step video:

Resources

Here are a few reference sites:

ShareGate MCP

ShareGate Protect

Agent flows overview — Copilot Studio (Microsoft Learn)

Use agent flows with your agent (Microsoft Learn)

Introducing MCP in Copilot Studio