Edit

Share via

Connect to Dataverse with model context protocol (preview)

[This topic is pre-release documentation and is subject to change.]

The Model Context Protocol (MCP) is an open protocol that enables seamless integration between LLM applications and external data sources and tools. Microsoft Dataverse can act as an MCP server, providing intelligent access to tables and records to various MCP clients like Copilot Studio agents, VS Code GitHub Copilot, Claude desktop, and many others. This integration standardizes and streamlines the interaction between AI models and Dataverse data, making it more efficient and effective for developers to apply Dataverse's rich data capabilities within their AI-driven applications.

Once connected to the Dataverse MCP Server, you can choose from various tools in the Power Platform environment. These tools are: list tables, describe table, read data, create record, update record, list prompts, execute prompt, list knowledge sources, and retrieve knowledge.

This article explains how to set up and use the Dataverse MCP server with Microsoft Copilot Studio, Claude desktop, or Visual Studio Code (VS Code) GitHub Copilot as an MCP client. By following the steps in this article, you can interact with Dataverse, asking natural language questions like "show me my contacts" and receive answers based on stored data.

Important

  • This is a preview feature.
  • Preview features aren’t meant for production use and might have restricted functionality. These features are subject to supplemental terms of use, and are available before an official release so that customers can get early access and provide feedback.

Connect to Dataverse using an MCP server in Microsoft Copilot Studio

  1. Go to Power App and select your environment from the top right environment selector.
  2. From the left navigation pane, select Agents > Create new agent. If the item isn’t in the side panel pane, select …More and then select the item you want.
  3. Select Create.
  4. Scroll down to the Tools section and select + Add tool.
  5. Select Model Context Protocol, and then select Dataverse MCP Server.
    1. If there is no existing Dataverse connection, you're prompted to do so.
  6. Select Add to agent.

The individual tools available on this MCP server can be viewed and modified by selecting ... > Edit next to the Dataverse MCP Server tool.

You can now interact with the Dataverse MCP Server tool in the Test your agent chat pane. Try commands like "list tables in Dataverse," "describe table account," or "how many accounts do I have." Dataverse MCP in Copilot Studio

Connect to Dataverse using an MCP Server with Claude or GitHub Copilot in VS Code

Prerequisites

These are the prerequisites for using a Dataverse MCP Server with Claude or GitHub Copilot in VS Code:

  1. Create a Dataverse connection for the MCP configuration
  2. Install the Dataverse MCP server local proxy
  3. Get the tenant ID of your Dataverse environment

Connecting to Dataverse with an MCP Server in Copilot Studio doesn't require any of these prerequisites.

Create a Dataverse connection for the MCP configuration

  1. Go to Power Automate. If necessary, change to the correct environment by selecting it from the top right.
  2. Select Connections on the left navigation pane, and then select + New connection on the command bar.
  3. Type Dataverse in the search box, and then select the green colored Microsoft Dataverse connector. Dataverse connector
  4. Complete the instructions on your screen.
  5. Note the user name in the connection Name, this should be the same name that you used to create the environment earlier.
  6. Select the connection to open it and copy the entire URL from the browser and save it. You need this URL for Claude desktop and VS Code MCP configuration. Copy entire browser URL

Install the Dataverse MCP server local proxy

These steps install the Dataverse MCP server local proxy that is used by the MCP client, such as Claude desktop or VS Code GitHub Copilot.

  1. Install the .NET SDK 8.0 either from Downloads or with this PowerShell command.

    winget install Microsoft.DotNet.SDK.8

  2. In the Terminal window you opened earlier, run this command to install the Microsoft.PowerPlatform.Dataverse.MCP local proxy.

    dotnet tool install --global --add-source https://api.nuget.org/v3/index.json Microsoft.PowerPlatform.Dataverse.MCP

Get the tenant ID of your Dataverse environment

When you configure the Dataverse MCP server for either Claude Desktop or VS Code GitHub, you need to provide the TenantID value.

Here’s one of the ways to get tenant ID details:

  1. Go to https://make.powerapps.com.
  2. Select Settings (gear icon) on the top right, and then select Session details.
  3. Copy the value of the Tenant ID from the Power Apps session details. Make a note of this GUID because it's used in the configuration steps later.

Configure and use the Dataverse MCP server in Claude

Claude AI is a large language model (LLM) and chatbot developed by Anthropic. It excels at natural language processing and is multimodal, meaning it can process text, audio, and visual inputs. Claude can answer questions, summarize documents, generate text, and even create diagrams, animations, and code.

Download Claude desktop

If you haven't already done so, download and install Claude desktop Download - Claude.

Once you have Claude desktop installed, you can find and launch Claude from your desktop.

Configure Dataverse MCP server in Claude desktop

  1. Open Claude desktop and go to File > Settings.

  2. If you haven't configured any MCP servers for Claude desktop previously, you observe a Settings dialog. Select Edit Config.

  3. This takes you to the Claude desktop files. Open the claude_desktop_config.json file with your favorite JSON editor.

  4. Replace <connection URL> and <Tenant Id> with your connection URL and tenant ID. More information: Create a Dataverse connection for the MCP configuration and Get the tenant ID of your Dataverse environment

    Use a <friendly name> for your Dataverse MCP server that you can easily remember, for example: MyDataverseMCPServer.

{
     "mcpServers": {
    "<friendly name>": {
      "command": "Microsoft.PowerPlatform.Dataverse.MCP",
      "args": [
        "--ConnectionUrl",
        "<URL for Dataverse connection>",
        "--MCPServerName",
        "DataverseMCPServer",
        "--TenantId",
        "<Tenant Id GUID>",
        "--EnableHttpLogging",
        "true",
        "--EnableMsalLogging",
        "false",
        "--Debug",
        "false",
        "--BackendProtocol",
        "HTTP"
         ]
       }
     }
   }
  1. Save this file and go back to Claude desktop.
  2. To restart Claude desktop and ensure that the changes take effect, select File > Exit.
  3. Open Claude desktop now that the Dataverse MCP server configuration is completed from the previous step. You need to use your credentials to sign in to your Dataverse environment.
  4. Verify that you can view the Dataverse MCP server and the tools by selecting Search and tools. You should be able to observe your friendly name of Dataverse MCP Server, MyDataverseMCPServer for example.

Claude connected to Dataverse 9. Selecting the MCP server (MyDataverseMCPServer) allows you to view the list of tools, supported by that MCP server.

Tip

You can enable and disable individual tools for each MCP server registered with Claude. This gives you control over what tools to use.

Interact with Dataverse MCP server in Claude desktop

If you have data in the Dataverse environment, you can start testing your setup by asking “list tables in Dataverse,” “describe table account,” or “how many accounts do I have,” and so on. More information: Add and remove sample data

Tip

If you have other MCP servers registered with Claude, it’s best to add in Dataverse in your prompt to be specific about which MCP server you’d like to use.

Configure and use your MCP client with VS Code

This section shows you how to configure your MCP server in two ways in VS Code GitHub Copilot: reusing Dataverse MCP server configuration that you have defined for Claude desktop or creating a new Dataverse MCP configuration for your VS Code GitHub Copilot.

If you don’t have VS Code installed, download Visual Studio Code - Mac, Linux, Windows.

Reuse Claude Desktop configuration

  1. In VS Code, open the command palette using Ctrl+Shift+P or View > Command Palette. Type MCP: and a list of relevant MCP commands are displayed, such as MCP: List Servers and MCP: Add Servers.

  2. If you have configured the Dataverse MCP server as described earlier in the Claude desktop step and your VS Code MCP setting is set as "chat.mcp.discovery.enabled": true, VS Code is able to discover it. For example, when you choose MCP: List Servers, the MCP server, such as MyDataverseMCPServer Running, is displayed. MCP Dataverse server running

    If the Dataverse MCP server isn't running, select the server and then select Start Server. Observe the server start in the Output window of VS Code.

Configure the Dataverse MCP server in VS Code

These instructions help you configure a Dataverse MCP server at the user setting level.

  1. In VS Code, go to Manage (gear on lower left) > Settings or CTRL+, and then type MCP.

  2. Mcp is listed under the User tab. Select Edit in settings.json. Edit Mcp JSON in VS Code

  3. Add the Dataverse MCP configuration text inside the mcp "servers" setting following the curly brace. JSON snippet location for MCP Dataverse 

    "<friendly name>": {
"command": "Microsoft.PowerPlatform.Dataverse.MCP",
"args": [
  "--ConnectionUrl",
  "<URL you copied from the screen with connection at make.powerautomate.com>",
  "--MCPServerName",
  "DataverseMCPServer",
  "--TenantId",
  "<GUID you copied from Tenant ID value of Settings at make.powerapps.com>",
  "--EnableHttpLogging",
  "true",
  "--EnableMsalLogging",
  "false",
  "--Debug",
  "false",
  "--BackendProtocol",
  "HTTP"
 ]
}

  4. Replace <connection URL> and <tenant ID> from the prerequisite steps. Use a <friendly name> for your Dataverse MCP server that you can easily remember, for example: MyDataverseMCPServerForGitHubCopilot.

Note

When the MCP server is configured correctly in settings.json, you notice a status like Start. This means that syntactically, it's correct and you can start the MCP server. In case it doesn’t show Start, you can go to Command Palette (Ctrl+Shift+P), type MCP: and then select MCP: List Servers. You should observe the friendly name that you have assigned for the Dataverse MCP server so you can start the MCP server.

Interact with Dataverse MCP server in VS Code GitHub Copilot

  1. In VS Code, open GitHub Copilot in Agent mode. Use CTRL+ALT+I to launch GitHub Copilot chat in VS Code.
  2. From this point on, you can interact with the MCP server via Agent mode of GitHub Copilot. For example, “list tables in Dataverse,” “describe table account,” or “how many accounts do I have,” and so on.

Tip

If you have other MCP servers registered with GitHub Copilot, there are a few ways to help MCP Client to choose the apporpriate MCP Server. Examples:

  • add "in Dataverse" to your prompt to be specific about which MCP server you’d like to use for your question, or
  • at the beginning of the session, you can say "Use <insert your MCP server name> for this session".

For more resources about how to use GitHub Copilot in VS Code:

To learn how to use Visual Studio Code and MCP Severs go to this document: Use MCP servers in VS Code (Preview)

List of tools available in Dataverse MCP server

The following Dataverse MCP tools are available. Your prompt in the MCP client like Claude desktop and VS Code GitHub Copilot is automatically routed to one or more of these tools. So you can ask a question like "view Accounts data," which is likely be mapped to the read_query tool or retrieve_knowledge.

Tool Description
create_record Insert a row into a table in Dataverse and returns the GUID of the created row.
describe_table Get the table schema of the requested table in Dataverse.
execute_prompt Execute a prompt from the list of available predefined prompts in the environment.
list_knowledge_sources Returns a list of knowledge sources available in Dataverse. Knowledge sources created and used in Copilot Studio agents in the same environment are shown here.
list_prompts List predefined prompts available in the environment.
list_tables List tables that are available in the environment
read_query Read data from tables in Dataverse.
retrieve_knowledge Use a preconfigured knowledge source to answer questions.
update_record Update a row in a Dataverse table.

Learn more about MCP: