Building the agent

A Copilot Studio maker or admin in your Microsoft 365 tenant. Your Cloverleaf organization admin also needs to have MCP enabled first — see Step 1 below.

Using the agent

Anyone with a Microsoft 365 Copilot license, once the agent has been published to your organization. Nothing to set up beyond signing in the first time.

Copilot is a general-purpose AI assistant with no knowledge of how you or your teammates think, communicate, or work best. Connecting the Cloverleaf Model Context Protocol (MCP) server gives Copilot access to your real assessment results and coaching insights — DISC, 16 Types, Motivating Values, and more.

That means coaching guidance you get inside Copilot is grounded in actual behavioral data rather than generic advice. Once connected, you can ask Copilot to help you prep for a 1:1, navigate a difficult conversation, or understand how two people work together, and get responses informed by your Cloverleaf Relationship Profiles.

Unlike Claude and ChatGPT, Microsoft doesn't have a public directory of MCP connectors yet, so Cloverleaf isn't something you pick from a list — a Copilot Studio maker adds it to an agent directly using its server URL, the same way you would in Claude Code or another MCP client.

Before anyone can connect Copilot to Cloverleaf, a Cloverleaf organization admin needs to switch on the MCP integration. This is a one-time setting that applies to your whole organization, and it's separate from the Copilot Studio setup below.

Only organization admins can see and change this setting. If you don't have access, ask a Cloverleaf admin in your organization to turn it on.

Copilot Studio needs an explicit Client ID and Client secret for Cloverleaf — unlike Claude and ChatGPT, it doesn't register a client automatically. Request OAuth credentials for your Copilot Studio agent from your Cloverleaf account team before you start.

The Copilot Studio agent builder — open an existing agent or create a new one, then find Tools in the right-hand panel.

The Add a tool dialog, filtered to the Model Context Protocol (MCP) category.

Server name, description, URL, and authentication set on the Add MCP server dialog.

Scrolled down: Client ID, Client secret, and the OAuth URLs and scopes.

Use Manual rather than Dynamic (with discovery) — dynamic discovery isn't reliable for Cloverleaf yet and can hang without connecting.

Naming and creating the connection for the Cloverleaf MCP server.

The Cloverleaf tool showing Ready in Copilot Studio's Tools list.

Publishing to the Teams + Microsoft 365 channel, with availability in Microsoft 365 Copilot turned on.

It can take a few hours for the agent to show up for everyone in your organization after publishing, so don't be surprised if it isn't visible to teammates right away.

You'll know the agent is set up when — the Cloverleaf tool shows Ready under Tools in the agent builder, and the agent's publish status shows Published with the Teams + Microsoft 365 channel turned on.

Some Copilot Studio environments haven't rolled out native MCP tools yet. If you don't see Model Context Protocol (MCP) as a category in the Add a tool dialog, you can add Cloverleaf as a Custom Connector instead, using a Generic OAuth 2 connection with the same server URL and OAuth values from the table above. Reach out to your Cloverleaf account team if you'd like help setting this up.

Once your organization's agent is published, using it takes about a minute. Sign-in only happens once — after that, Copilot remembers you.

Your organization's agent may be named differently — at Cloverleaf, our internal agent is called Cloverleaf Internal. Swap in your own agent's name below.

Searching “Cloverleaf” in the Agent Store — your organization may have more than one Cloverleaf agent listed.

The agent's landing page, with suggested conversation starters like “My daily coaching,” “Ask your coach,” and “DISC profile.”

Type a question, or click one of the suggested prompts (for example, DISC profile → “Render my DISC profile”). Copilot shows what it's about to do and asks you to confirm before it calls a Cloverleaf tool for the first time in a conversation — click Confirm.

Copilot asking to confirm the tool call — here, rendering a DISC profile widget — before it runs.

The first time you use a Cloverleaf tool, Copilot asks you to sign in.

Copilot prompting you to sign in before it can process the request.

The Cloverleaf authorization window — review the permissions, then click Authorize.

Real DISC score data returned after signing in — scores, style, and what it means, pulled from your actual Cloverleaf assessment.

With the connector active, you can ask Copilot questions like:

You, in Copilot“What does my Cloverleaf DISC profile say about how I show up under pressure?”

You, in Copilot“Give me my coaching for today.”

You, in Copilot“Help me prep for a 1:1 with [teammate name].”

You, in Copilot“How should I give [teammate name] feedback about a missed deadline so it actually lands?”

You, in Copilot“Show me the relationship dynamics between me and [teammate name].”

Copilot will pull your real assessment data and interpreted personality dynamics directly into the conversation.

If none of these fix it, contact your Cloverleaf account team or your internal IT/Copilot Studio admin with a screenshot of what you see.