Set up the remote Azure DevOps MCP Server (preview)

Azure DevOps Services

The remote Azure DevOps MCP Server is a hosted version of the Azure DevOps MCP Server that doesn't require a local installation. Instead of running the server on your machine, you connect your AI assistant directly to the Azure DevOps–hosted endpoint by using streamable HTTP transport.

The remote server provides the same capabilities as the local server, including access to work items, pull requests, pipelines, and more, while eliminating local setup complexity.

Remote vs. local MCP Server

Prerequisites

mcp.json configuration

{
  "servers": {
    "ado-remote-mcp": {
      "url": "https://mcp.dev.azure.com/{organization}",
      "type": "http"
    }
  },
  "inputs": []
}

You can also omit the organization name from the URL. However, if you omit the organization name, you must provide it as context in each tool call.

{
  "servers": {
    "ado-remote-mcp": {
      "url": "https://mcp.dev.azure.com/",
      "type": "http"
    }
  },
  "inputs": []
}

Toolsets

Specify toolsets to restrict the tools available to the MCP server. Should not be combined with X-MCP-Tools.

{
  "servers": {
    "ado-remote-mcp": {
      "url": "https://mcp.dev.azure.com/{organization}",
      "type": "http",
      "headers": {
        "X-MCP-Toolsets": "repos,wiki,wit"
      }
    }
  },
  "inputs": []
}

Read-only tools

Use the X-MCP-Readonly header to restrict the server to read-only operations. This restriction is useful when you want to prevent modifications to Azure DevOps resources.

{
  "servers": {
    "ado-remote-mcp": {
      "url": "https://mcp.dev.azure.com/{organization}",
      "type": "http",
      "headers": {
        "X-MCP-Readonly": "true"
      }
    }
  },
  "inputs": []
}

Combine toolsets and read-only filtering

{
  "servers": {
    "ado-remote-mcp": {
      "url": "https://mcp.dev.azure.com/{organization}",
      "type": "http",
      "headers": {
        "X-MCP-Toolsets": "repos,wiki,wit",
        "X-MCP-Readonly": "true"
      }
    }
  },
  "inputs": []
}

Individual tools

Use the X-MCP-Tools header to enable only specific tools. Should not be combined with X-MCP-Toolsets.

{
  "servers": {
    "ado-remote-mcp": {
      "url": "https://mcp.dev.azure.com/{organization}",
      "type": "http",
      "headers": {
        "X-MCP-Tools": "core_list_projects, wit_my_work_items, wit_get_work_items_batch_by_ids"
      }
    }
  },
  "inputs": []
}

Insiders

As we experiment and introduce new tools and updates to existing ones, you can get early access to these changes by using the X-MCP-Insiders header.

{
  "servers": {
    "ado-remote-mcp": {
      "url": "https://mcp.dev.azure.com/{organization}",
      "type": "http",
      "headers": {
        "X-MCP-Insiders": "true"
      }
    }
  },
  "inputs": []
}

Available tools

Core tools

Core tools are always available.

Work

Repos

The repository tools are consolidated into grouped dispatchers using an action parameter, following the same pattern as the wiki tools. The previous individual tool names continue to work as aliases. Use repo_create_branch and repo_search_commits directly — they were not consolidated.

Wit

The work item tools are consolidated into grouped dispatchers using an action parameter, following the same pattern as the repo and pipeline tools. The previous individual tool names continue to work as aliases. Use search_workitem directly — it was not consolidated.

Pipelines

The pipeline tools are consolidated into grouped dispatchers using an action parameter, following the same pattern as the wiki and repo tools. The previous individual tool names continue to work as aliases.

Wiki

The wiki read operations are consolidated into a single wiki tool. Use the action parameter to select the operation.

Test plans

Search

Supported environments

The remote Azure DevOps MCP Server requires your user account and Azure DevOps organization to connect to Microsoft Entra ID.

Not all MCP clients support Entra authentication by default. Some environments require extra steps to register the client application.

Currently supported environments include:

• Visual Studio Code
• Visual Studio

Visual Studio Code

To configure the remote MCP Server in Visual Studio Code:

1. Add a .vscode/mcp.json file to your repository with the mcp.json configuration shown earlier.
2. Save the file.
3. Open GitHub Copilot in VS Code.
4. When prompted, authenticate by using your Microsoft Entra account and select your account.

After authentication completes, a list of available tools appears.

Visual Studio (2022 and later)

Configure the remote MCP Server in Visual Studio by adding the server URL to your MCP settings. For more information, see Use MCP servers in Visual Studio.

Verify the connection

After setup, verify the remote MCP Server is working by asking your AI assistant a question about your Azure DevOps data.

Examples:

• "List the projects in my Azure DevOps organization."
• "Show my assigned work items."
• "What pull requests require my review?"

If the AI assistant returns results from your Azure DevOps organization, the remote MCP Server is configured correctly.

Troubleshooting

For support, you can create an issue in the local MCP Server repo. Be sure to use the Remote issue template.

FAQ

What about other clients like GitHub Copilot CLI, Claude Desktop, Claude Code, CodeX, or Cursor?

Additional client tools such as CodeX, Claude Desktop, Claude Code, and ChatGPT require dynamic registration of an OAuth Client ID in Entra before they can be used with the MCP server. We are working closely with the Entra team to enable this capability. For now, only Visual Studio and Visual Studio Code are supported.

What services can use Azure DevOps MCP Server?

Support for other services, including Azure AI Foundry, Microsoft 365 Copilot, and Copilot Studio, is not yet available but will be added soon.

Does the Azure DevOps MCP Server support AgentId?

AgentId support is coming soon.

Related content

• Azure DevOps MCP Server overview
• Azure DevOps MCP Server GitHub repository