Use Azure Databricks data in Microsoft Copilot Studio

This page explains how to use your Azure Databricks data in Microsoft Copilot Studio after creating a connection. You can connect Genie Spaces as tools or use your Azure Databricks data as a knowledge source to build custom agents.

Prerequisites

Before you use Azure Databricks data in Copilot Studio, you must create a connection to Azure Databricks on Microsoft Power Platform. See Create an Azure Databricks connection on Microsoft Power Platform.

Connect to Genie Spaces in Copilot Studio

Important

This feature is in Public Preview.

Important

Enable the Managed MCP Servers preview in your workspace. See Manage Azure Databricks previews.

Genie is a Azure Databricks feature that allows business teams to interact with their data using natural language. For more information about setting up Genie Spaces on Azure Databricks, see Genie Spaces. Make sure to provide your Azure Databricks Genie Space a clear name and description.

To add a Genie Space as a tool in a Copilot Studio agent, do the following:

  1. From the Copilot Studio sidebar, click Agent.

  2. Select an existing agent or create a new agent by clicking + Create blank agent.

  3. Select your agent's model. Databricks recommends selecting Sonnet 4.5 as your agent's model.

  4. Under Instructions, describe your agent's capabilities and how it should behave.

    • Databricks recommends including the following in the instructions: “Always continue polling the Genie Space until you receive a response. Do not stop after a few seconds.” This prevents the integration from repeatedly sending a "still processing" message while the agent is polling the Genie Space.
  5. Under Tools, click + Add a Tool.

  6. Search for "Azure Databricks" or select Model Context Protocol.

  7. Select Azure Databricks Genie, choose your connection to Azure Databricks next to Connection, and click Add and configure.

  8. Configure the Azure Databricks Genie tool:

    1. Name the tool. A more descriptive tool name helps the Copilot Studio agent orchestrate requests.
    2. Under Inputs, enter the Genie Space ID or select the Genie Space from the dropdown menu.
    1. To use on-behalf-of (OBO) flow, under Details, click Additional details. Select End user credentials as the Credentials to use. To use a singular identity, select Maker-provided credentials.
    1. Save the tool.
    2. (Optional) Refresh the tools section in the configuration screen to confirm that the Genie Space is connected.
  9. In the upper-right, click Settings. Under the Orchestration section, click Yes to use generative AI orchestration for your agent's responses.

  10. To publish your agent, click Publish in the upper-right side.

  11. (Optional) Publish the Genie-enabled Copilot Studio agent to Microsoft Teams or Microsoft 365 to distribute Genie's insights to other users. To learn how to publish your agent, see Microsoft's documentation.

You can add your agent to a Power App. To set up a Power App, see Use your Azure Databricks data to build Power canvas apps. To add your Copilot Studio agent to a Power App, see Microsoft's documentation.

To use the agent in your published app, open the published application, click the Copilot icon in the upper-right corner, and ask a question.

Use Azure Databricks as a knowledge source in Copilot Studio

To add your Azure Databricks data as a knowledge source to a Copilot Studio agent, do the following:

  1. From the Copilot Studio sidebar, click Agent.

  2. Select an existing agent or create a new agent by clicking + New agent.

    • Describe the agent by inputting a message and then click Create.
    • Or, click Skip to manually specify the agent’s information.
  3. On the Knowledge tab, click + Knowledge.

  4. Click Advanced.

  5. Click Azure Databricks as the knowledge source.

  6. Enter the catalog name your data is in.

  7. Click Connect.

  8. Select the tables you want your agent to use as a knowledge source and click Add.

Limitations

  • Genie Spaces in Copilot Studio support up to five questions per minute due to Genie API rate limits. Free tier limits are in place to prevent abuse. To scale beyond, contact your Databricks account team.