ContextQMD
Back to Datahub

Microsoft Copilot Studio Integration

docs/dev-guides/agent-context/copilot-studio.md

1.5.0.36.3 KB
Original Source

Microsoft Copilot Studio Integration

📚 Navigation: ← Back to Agent Context Kit | LangChain Integration → | Snowflake Integration →

What Problem Does This Solve?

Microsoft Copilot Studio lets you build custom AI agents that work across Microsoft 365, Teams, and other channels. Out of the box, these agents don't know anything about your organization's data assets — they can't search your data catalog, trace lineage, or answer questions about dataset ownership and quality.

By connecting DataHub's MCP server to Copilot Studio, your agents gain direct access to your metadata catalog. Users can ask questions like "What are the most queried datasets?" or "Who owns the revenue table?" and get grounded answers drawn from real metadata.

Prerequisites

Setup Guide

Step 1: Create or Open an Agent

  1. Navigate to Copilot Studio and sign in.
  2. Create a new agent by clicking + Create a blank agent, or open an existing one.
<p align="center"> </p>

Step 2: Add an MCP Tool

  1. From your agent's overview page, scroll down to the Tools section and click + Add tool.
<p align="center"> </p>
  1. In the "Add tool" dialog, select Model Context Protocol under "Create new".
<p align="center"> </p>

Step 3: Configure the MCP Server Connection

Fill in the MCP server details:

FieldValue
Server nameDataHub MCP Server
Server descriptionDataHub MCP server to connect to datasets, documents, & more
Server URLhttps://<tenant>.acryl.io/integrations/ai/mcp
AuthenticationAPI key
TypeHeader
Header nameAuthorization
Header valueBearer <your-datahub-token>

Replace <tenant> with your DataHub Cloud tenant name and <your-datahub-token> with your personal access token.

<p align="center"> </p>

:::note Self-Hosted DataHub If you're using a self-hosted DataHub instance, you'll need to run the self-hosted MCP server and expose it via a publicly accessible URL. Use that URL as the Server URL above. :::

Step 4: Add and Configure Tools

  1. After entering the connection details, click Next. Copilot Studio will discover the available tools from the DataHub MCP server.
  2. Review the connection and click Add and configure to add the MCP server to your agent.
<p align="center"> </p>
  1. You should now see all available DataHub tools listed. Toggle on the tools you want your agent to have access to.
<p align="center"> </p>

Step 5: Test Your Agent

Click the Test button in the top-right corner to open the test panel. Try asking your agent a question like:

  • "Find the most important data at my organization"
  • "What datasets does the analytics team own?"
  • "Show me the lineage for the revenue dashboard"
<p align="center"> </p>

Tips

  • Add instructions: Use the Instructions field on the agent overview page to guide how the agent uses DataHub tools — e.g., "Always search DataHub before answering data questions" or "Include dataset owners in your responses."
  • Add knowledge: You can attach additional documents or data sources under the Knowledge section to complement DataHub metadata.
  • Publish your agent: Once you're happy with the agent, click Publish to make it available in Teams, your website, or other channels.

Troubleshooting

Connection Errors

If Copilot Studio can't connect to the MCP server:

  • Verify your DataHub Cloud instance URL is correct (https://<tenant>.acryl.io/integrations/ai/mcp)
  • Ensure the personal access token is valid and hasn't expired
  • Confirm that authentication is set to API key (not OAuth 2.0)
  • Make sure the header value includes the Bearer prefix (with the trailing space)

Tools Not Appearing

If tools don't appear after adding the MCP server:

  • Click the refresh button on the Tools page to re-discover tools
  • Verify the MCP server is running and healthy
  • Check that your token has the required permissions

Empty Results

If the agent returns no results for queries: