Azure Storage MCP Apps Demo
Provides an interactive dashboard within VS Code Copilot to manage Azure Storage accounts, containers, and blobs through a rich UI. It enables users to visualize storage resources, perform CRUD operations, and generate SAS tokens using the Model Context Protocol.
README
Azure Storage MCP Apps Demo
An interactive MCP Apps demo that renders an Azure Storage dashboard directly inside VS Code Copilot chat. Browse storage accounts, manage containers and blobs, and generate SAS tokens -- all within a rich UI powered by the MCP Apps Extension.
What are MCP Apps?
MCP Apps extend the Model Context Protocol by letting MCP tools return interactive UI components (HTML/CSS/JS) that render directly in the conversation. Instead of plain text responses, tools can display dashboards, forms, visualizations, and multi-step workflows -- all while keeping the AI model informed of user actions.
Demo Flow
- Ask Copilot: "Show me all Azure Storage Accounts in my subscription"
- Dashboard appears: An interactive card grid showing storage accounts with location, SKU, tier, and size info
- Click a storage account: The dashboard calls
get_storage_accountand renders a detail panel with properties, endpoints, and a containers table - Manage containers: Create new containers, delete existing ones, or click into a container to browse blobs
- Manage blobs: Upload files (drag-and-drop or text), download blobs, preview content inline, or delete blobs
- Generate SAS tokens: A form modal lets you select permissions and expiry; the UI calls
generate_sas_tokenand displays the token and full URL with copy buttons
All interactions happen within the embedded UI -- the agent orchestrates tool calls behind the scenes.
Prerequisites
- Node.js 18+
- VS Code with GitHub Copilot (Copilot Chat)
- MCP Apps support in your VS Code version (VS Code Insiders recommended)
- Azure subscription with the
AZURE_SUBSCRIPTION_IDenvironment variable set - Azure CLI logged in (or another credential source supported by
DefaultAzureCredential)
Quick Start
# Install dependencies
npm install
# Build (type-check + bundle UI + compile server)
npm run build
# Configure the MCP server in VS Code (see .vscode/mcp.json example below)
# Then open this folder in VS Code and use Copilot chat
Development
# Watch mode: auto-rebuilds UI and restarts server on changes
npm run dev
# Run HTTP server for testing (port 3001)
npm run serve
# Run stdio server directly
npm run serve:stdio
Project Structure
├── mcp-app.html # UI entry point (bundled by Vite)
├── src/
│ ├── mcp-app.ts # Client-side dashboard logic (App class)
│ └── mcp-app.css # Azure-themed styling
├── server.ts # MCP server: 10 tools + 1 UI resource
├── main.ts # Entry point: stdio + HTTP transports
├── dist/ # Build output (bundled HTML + compiled JS)
├── package.json
├── tsconfig.json # Type checking config
├── tsconfig.server.json # Server compilation config
└── vite.config.ts # Vite + singlefile bundler
Architecture
MCP Server (server.ts)
Registers ten tools and one UI resource using @modelcontextprotocol/ext-apps/server:
| Tool | Description | Has UI? |
|---|---|---|
list_storage_accounts |
Lists all accounts in subscription | Yes (dashboard) |
get_storage_account |
Returns detailed account info | No (consumed by dashboard UI) |
list_containers |
Lists blob containers in an account | No (consumed by dashboard UI) |
create_container |
Creates a new blob container | No (consumed by dashboard UI) |
delete_container |
Deletes a blob container | No (consumed by dashboard UI) |
list_blobs |
Lists blobs in a container | No (consumed by dashboard UI) |
upload_blob |
Uploads content as a blob | No (consumed by dashboard UI) |
download_blob |
Downloads blob content | No (consumed by dashboard UI) |
delete_blob |
Deletes a blob | No (consumed by dashboard UI) |
generate_sas_token |
Generates a SAS token for a container | No (consumed by dashboard UI) |
Dashboard UI (src/mcp-app.ts)
Uses the App class from @modelcontextprotocol/ext-apps to:
- Receive initial tool results via
ontoolresult - Navigate between views (accounts → containers → blobs) with breadcrumb navigation
- Call server tools via
app.callServerTool()for all CRUD operations (get details, list/create/delete containers, list/upload/download/delete blobs, generate SAS) - Display modal dialogs for create container, upload blob, generate SAS token, confirm delete, and blob content preview
- Apply host theming via
onhostcontextchanged
Build Pipeline
- Vite + vite-plugin-singlefile bundles
mcp-app.html+src/mcp-app.ts+src/mcp-app.cssinto a single self-contained HTML file indist/ - TypeScript compiles server files (
server.ts,main.ts) todist/ - The MCP server reads and serves the bundled HTML as a
text/html;profile=mcp-appresource
References
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。