|
This article covers the following: |
Overview
You can connect your Wingify account to a Microsoft Copilot Studio agent through the Wingify Model Context Protocol (MCP) server. This integration allows the agent to answer questions about your campaigns, heatmaps, and session recordings in natural language. For example, you can enter the prompt: List my Wingify campaigns that are currently running. The agent fetches all the relevant campaign details.
The Wingify MCP server acts as a secure bridge between external AI tools and your live Wingify data. It operates separately from Wingz. Wingz is the AI assistant within the Wingify application, while the MCP server is the backend that external tools, such as Copilot Studio, connect to.
For example, once you connect the Wingify MCP server to the agent in Copilot Studio, a marketer can ask the agent in Microsoft Teams, Which of my Wingify campaigns running in the last 30 days have no winner yet?
The agent calls the Wingify MCP server and returns the list without opening the Wingify dashboard.
The Wingify MCP integration enables you to:
- Query live campaign performance, heatmap insights, and session recording summaries using natural language.
- Create, clone, or manage A/B test and Split URL campaigns directly from your Copilot Studio agent (based on granted permissions).
- Control granular access by assigning specific permission levels and enabling or disabling individual MCP tools.
Prerequisites
Before you connect Wingify to Microsoft Copilot Studio, ensure you have the following:
- A Wingify account and a Wingify user who can approve access for the account you want to connect.
- Access to Microsoft Copilot Studio with permission to create or edit an agent in your environment. Microsoft sets licensing requirements, confirm your license status with your Microsoft 365 administrator.
- Generative orchestration enabled for your agent.
- If your organization uses Power Platform data policies, an administrator must allow custom connectors. Copilot Studio connects to MCP servers through Power Platform connectors, so those policies also govern Wingify access.
-
Identify and use the Wingify MCP server URL that corresponds to your account region:
Region MCP Server URL Global https://mcp.wingify.ai/mcp Europe https://mcp-eu.wingify.ai/mcp Asia https://mcp-as.wingify.ai/mcp
Configure and Enable Wingify MCP Server in Copilot Studio
Follow these five steps to add the Wingify MCP server, authorize your account, and configure tool permissions in Microsoft Copilot Studio.
Step 1: Add the Wingify MCP Server to Your Agent
To add the Wingify MCP server to the Copilot Studio agent:
- Log in to Microsoft Copilot Studio and open an existing agent, or create a new one.
- Go to the agent's Tools page.
- Select Add a tool > New tool.
- Select Model Context Protocol. The MCP onboarding wizard opens.
-
Enter the following server details:
- Server name: Enter a name for the server, for example, Wingify MCP.
- Server description: Describe what the server does. For example, enter Reads and manages Wingify A/B test and Split URL campaigns, campaign reports, heatmaps, and session recordings. Copilot Studio uses this description to decide when the agent calls Wingify, so keep it clear and specific.
-
Server URL: Enter the MCP server URL for your region for example, https://mcp.wingify.ai/mcp
For more information, refer to Connect your agent to an existing Model Context Protocol (MCP) server.
- Under authentication, select OAuth 2.0, and then set Type to Dynamic discovery.
-
Click Create, and then click Next. The Add tool dialog opens.
Step 2 explains how you can set up the connection in the Add tool dialog.Note: Copilot Studio supports only the Streamable HTTP transport for MCP. Server-Sent Events (SSE) is no longer supported. If the wizard fails to connect to the Wingify server, see Troubleshooting.
Step 2: Create MCP Connection and Authorize Wingify
The Wingify MCP connection authorizes the agent to access your Wingify account via OAuth. To configure and authorize the Wingify MCP connection with the Copilot agent, follow these steps:
- In the Add tool dialog, select Create a new connection, or choose an existing Wingify connection.
- When the Wingify authorization screen opens, select the Wingify account from the Select an account dropdown.
-
From the Permission level dropdown, choose what the agent can do:
Permission Description Browse Provides read-only access to Wingify data, session recordings, heatmaps, A/B tests, Split URL tests, and Multivariate tests. Cannot create or edit campaigns. Design Enables creation of new A/B or Split URL tests via the MCP server. Publish / Admin Provides full control to create, edit, start, pause, or delete A/B tests and Split URL tests. - Select the acknowledgment checkbox and click Approve.
- In Copilot Studio, click Add to agent (or Add and configure). Wingify MCP now appears in the agent's list of tools.
Step 3: Choose which Wingify Tools the Agent Can Use
By default, the agent can use all Wingify tools your permission level allows. To restrict which tools the agent can access:
- In Copilot Studio, on the agent's Tools page, select Wingify MCP. The Wingify MCP Settings page opens and lists all available tools with their descriptions.
- Toggle off the Allow all switch. An individual toggle switch appears beside each tool.
- Toggle off the tools the agent does not need. For example, toggle off all Write tools for a reporting-only agent.
- Save your changes.
Step 4: Test the Agent and Query Your Wingify Data
Verify that the integration works correctly by testing the agent in Copilot Studio before sharing it with your team. To test the agent for Wingify-related queries:
- Open the Test agent panel in Copilot Studio.
- Enter a Wingify question in natural language. The first time you query the agent, it may prompt you to sign in to the Wingify connection. Complete the sign-in to continue.
-
Confirm that the agent calls a Wingify tool and returns live data from your account.
You can use the following sample prompts to test your agent:
- List my Wingify campaigns that are currently running.
- Which campaigns published in the last seven days have no winner yet?
- Show the daily data for the campaign Pricing Page Headline for the last 14 days.
- Analyze the heatmap for https://example.com/pricing and summarize where visitors drop off.
- What sample size do I need to detect a 5 percent lift on a 3 percent conversion rate?
Step 5: Publish the Agent to Teams and Microsoft 365 Copilot
Once testing succeeds, publish the agent to make it available to your team:
- With the agent open in Copilot Studio, click Publish, and then confirm. Publishing may take a few minutes.
- From the top menu, select Channels, then select Teams and Microsoft 365 Copilot.
-
Follow the configuration steps in Connect and configure an agent for Teams and Microsoft 365 Copilot.
Note:
- You can publish the agent across multiple channels or platforms, including live websites, mobile apps, Microsoft Copilot, and messaging platforms such as Teams and Facebook. For more information, see Key concepts - Publish and deploy your agent.
- A Copilot Studio agent is not auto-deployed to Teams and Microsoft 365 Copilot when published. Admins must approve it in the Teams Admin Center or Microsoft 365 Admin Center.
- Each user who interacts with the agent may be prompted to sign in to the Wingify connection. The agent can only perform actions permitted by the connection's permission level and the enabled tools.
Troubleshooting
| Issue | Possible Cause | Recommended Solution |
| Model Context Protocol is missing under Add a tool > New tool. | Your environment or license does not allow MCP tools, or a Power Platform data policy blocks custom connectors. | Ask your Power Platform administrator to check your Copilot Studio license and data policies. |
| Selecting the Model Context Protocol connector opens a docs page instead of the wizard. | The MCP Wizard feature is not yet enabled in your tenant/environment. | Check with the support team to verify whether the Add existing MCP server / MCP config wizard feature flag is enabled for your environment. |
| The wizard fails to connect, or the Wingify authorization screen fails to open. | You entered the wrong regional server URL, or the authentication type is not set to OAuth 2.0 > Dynamic discovery. | Re-enter the server URL for your account's region and select OAuth 2.0 > Dynamic discovery. |
| The agent answers without fetching details from Wingify. | The server description is too vague for the agent to know when to use Wingify. | Make the Server description more specific, or include Wingify and a specific campaign name in your prompt. |
| The agent cannot create, start, or pause a campaign. | The connection uses Browse permission, or the required Write tools are toggled off. | Re-create the connection with Design, Publish, or Admin permission, and toggle on the required Write tools in Step 3. |
| A new Wingify tool does not appear for the agent. | The Allow all toggle is switched off, so newly released tools are disabled by default. | Open the Wingify tool settings page and toggle on the new tool. |
FAQs
-
Is MCP integration the same as Wingz?
No. Wingz is the AI assistant available inside the Wingify application. This integration connects an external Microsoft Copilot Studio agent to the Wingify MCP server.
-
Can the agent create, update, or delete Multivariate tests?
No. Through the MCP server, AI tools can create and manage only A/B tests and Split URL campaigns. Multivariate tests are available in read-only mode.
-
Do I need an API key?
No. The Wingify MCP server uses OAuth 2.0 with Dynamic Client Registration, so no API key or client secret is required.
Need more help?
For more information or further assistance, contact Wingify Support.