学习如何将 Power BI 语义模型任务——文档、命名审计等——转到 GitHub Copilot 编码代理，由 Power BI 建模 MCP 服务器驱动。打开 GitHub 问题，将其分配给代理，并审查它交付的拉取请求。

工作原理

该模板包含三个部分：

1. Power BI 项目以 PBIP 格式呈现——语义模型和报表源代码，作为 Power BI 项目文件存储在 src/。
2. Power BI 建模 MCP 服务器 - 赋予编码代理对语义模型（表、测量、关系、RLS 等）的程序访问。
3. 技能——在.github/skills/中精心策划的逐步工作流程，指导客服人员产生一致且符合团队标准的结果。

开始

1. 创建你的仓库 - 点击使用此模板并选择在你的GitHub账户中创建新仓库。

2. 配置MCP服务器 - 导航到Copilot>云代理>设置，并粘贴以下内容到MCP配置中：

点击保存MCP配置。

3. 创建问题并分配代理 - 在 GitHub 上创建一个新问题，描述你希望在语义模型上执行的任务。

示例问题：

Agentic Power BI Development with GitHub Copilot Coding Agent

Learn how to Power BI semantic model tasks - documentation, naming audits, and more - to GitHub Copilot coding agent, powered by the Power BI Modeling MCP Server. Open a GitHub issue, assign it to the agent, and review the pull request it delivers.

How It Works

This template combines three pieces:

1. Power BI project in PBIP format - the semantic model and report source code, stored as Power BI Project files in src/.
2. Power BI Modeling MCP Server - gives the coding agent programmatic access to the semantic model (tables, measures, relationships, RLS, etc.).
3. Skills - curated step-by-step workflows in .github/skills/ that guide the agent to produce consistent, team-standard results.

Getting Started

1. Create your repository - Click Use this template and select Create a new repository in your GitHub account.

2. Configure the MCP server - Navigate to Settings > Copilot > Cloud Agent and paste the following into MCP configuration:

   {
       "mcpServers": {
           "powerbi-modeling-mcp": {
               "type": "stdio",
               "command": "npx",
               "args": [
                   "-y",
                   "@microsoft/powerbi-modeling-mcp",
                   "--start"
               ],
               "tools": ["*"]
           }
       }
   }
   

   Click Save MCP Configuration.

3. Create an issue and assign the agent - Open a new GitHub issue describing the task you want performed on the semantic model.

   Example issues:

   Generate documentation for the Sales semantic model

   Analyze the naming conventions of the Sales semantic model and propose changes

   Then assign the issue to the coding agent:

4. Review the pull request - Copilot creates a branch, executes the task, and opens a PR for your review:

   You can open the Copilot session log to verify that the skills and Power BI Modeling MCP were loaded successfully:

   The agent follows the instructions defined in the skills, so the output conforms to your team’s standards:

[!TIP]

• Try creating issues from your smartphone - you can delegate tasks to the coding agent from anywhere and focus on reviewing the results.
• This is a starting point. Customize the skills and add new ones to match your team’s development standards and workflows.