Azure Dev Server
A remote MCP server that provides full filesystem and shell access to a cloud-hosted environment via HTTPS, enabling Claude Desktop to read, write, edit, search files, and execute commands on a remote Azure machine.
README
⚠️ DEPRECATED — This repository has been consolidated into the
datacoremonorepo atpackages/dev-server. This repo is archived and no longer maintained. All future development happens in the monorepo.
azure-dev-server
Remote MCP server that gives Claude Desktop full filesystem and shell access over HTTPS — a cloud-hosted Desktop Commander running on Azure Container Apps.
What it does
Exposes 8 MCP tools so Claude Desktop can read, write, edit, search, and run commands on a remote machine:
| Tool | Description |
|---|---|
ping |
Health check |
read_file |
Read a file with line numbers |
write_file |
Write or append to a file |
edit_file |
Surgical find-and-replace (single occurrence) |
search_files |
Grep across a directory with optional file pattern |
list_directory |
List files/dirs with recursion depth |
run_command |
Execute a shell command, returns stdout/stderr |
get_file_info |
File metadata (size, type, modified time) |
Why it exists
Claude Desktop's built-in Desktop Commander only works locally. This project gives Claude Desktop the same capabilities on a remote Azure machine — useful when the workspace lives in the cloud or needs to be shared across devices.
Quickstart (local)
git clone https://github.com/david3xu/azure-dev-server.git
cd azure-dev-server
pnpm install
pnpm run build
MCP_API_KEY=your-key WORKSPACE=/path/to/your/workspace pnpm start
Health check:
curl http://localhost:3001/health
Connect to the live instance
The server runs at:
https://ca-dev-server.purplegrass-77b8c839.australiaeast.azurecontainerapps.io
Cursor
Add to ~/.cursor/mcp.json:
{
"mcpServers": {
"azure-dev": {
"url": "https://ca-dev-server.purplegrass-77b8c839.australiaeast.azurecontainerapps.io/mcp",
"headers": {
"x-api-key": "<your-api-key>"
}
}
}
}
Reload the window (Cmd+Shift+P → "Reload Window") — the 8 tools appear in the MCP panel.
VSCode + GitHub Copilot
Add to .vscode/mcp.json (per workspace) or user settings.json:
{
"mcp": {
"servers": {
"azure-dev": {
"type": "http",
"url": "https://ca-dev-server.purplegrass-77b8c839.australiaeast.azurecontainerapps.io/mcp",
"headers": {
"x-api-key": "<your-api-key>"
}
}
}
}
}
Claude Desktop
Claude Desktop only supports stdio transport, so use mcp-remote as an HTTP bridge.
Create ~/Library/Application Support/Claude/azure-dev-mcp-bridge.sh:
#!/bin/zsh
set -euo pipefail
exec /opt/homebrew/bin/npx -y mcp-remote \
"https://ca-dev-server.purplegrass-77b8c839.australiaeast.azurecontainerapps.io/mcp" \
--transport http-first \
--header "x-api-key:<your-api-key>"
Then add to ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"azure-dev": {
"command": "/bin/zsh",
"args": ["/Users/<you>/Library/Application Support/Claude/azure-dev-mcp-bridge.sh"]
}
}
}
Restart Claude Desktop (Cmd+Q, reopen).
Terminal (interactive shell inside the container)
Requires Azure CLI.
Step 1 — Add the alias to your shell profile (once):
echo "alias azure-dev-shell='az containerapp exec \\
--name ca-dev-server \\
--resource-group rg-datacore \\
--command bash'" >> ~/.zshrc
Step 2 — Load it in your current terminal:
source ~/.zshrc
Step 3 — Connect:
azure-dev-shell
You will see INFO: Successfully connected to container: 'ca-dev-server' and a shell prompt.
Tips inside the container:
clearis not available — useCtrl+Linstead- The prompt may show
sh-5.2#— runbashto switch to a full Bash session/workspaceis the Azure Files mount — files written here persist across restarts- Exit with
Ctrl+D
Verify the connection
After connecting via any client, run these tools in order:
ping— should return"alive — workspace: /workspace"list_directorywithpath: "/"— lists the workspace rootwrite_filethenread_file— confirms read/write access
Environment variables
| Variable | Required | Default | Description |
|---|---|---|---|
MCP_API_KEY |
Yes | dev-key-change-me |
API key sent in x-api-key header |
WORKSPACE |
No | process.cwd() |
Root directory for all file operations |
PORT |
No | 3001 |
HTTP port |
Deploy to Azure Container Apps
Requires Azure CLI. No local Docker needed — image is built in the cloud:
# One-time setup
az group create --name rg-datacore --location australiaeast
az acr create --name acrdevserver --resource-group rg-datacore --sku Basic --admin-enabled true
az containerapp env create --name cae-dev-server --resource-group rg-datacore --location australiaeast
# Build + deploy
az acr build --registry acrdevserver --image azure-dev-server:latest --file Dockerfile .
ACR_PASS=$(az acr credential show --name acrdevserver --query "passwords[0].value" -o tsv)
az containerapp create \
--name ca-dev-server \
--resource-group rg-datacore \
--environment cae-dev-server \
--image acrdevserver.azurecr.io/azure-dev-server:latest \
--registry-server acrdevserver.azurecr.io \
--registry-username acrdevserver \
--registry-password "$ACR_PASS" \
--target-port 3001 --ingress external \
--min-replicas 1 --max-replicas 3 \
--cpu 0.5 --memory 1.0Gi \
--env-vars "MCP_API_KEY=<your-key>" "NODE_ENV=production" "PORT=3001" "WORKSPACE=/workspace"
Re-deploy after code changes:
az acr build --registry acrdevserver --image azure-dev-server:latest --file Dockerfile .
az containerapp update --name ca-dev-server --resource-group rg-datacore \
--image acrdevserver.azurecr.io/azure-dev-server:latest
Development
pnpm run check # format + lint + type-check + build + test (all layers)
pnpm run test # tests only
pnpm run format # auto-format src/ and tests/
All 7 constraint layers active: CI → pre-commit hook → ESLint → TypeScript strict → tests → Prettier → Zod schemas.
Tech stack
@modelcontextprotocol/sdk— Streamable HTTP transport- Express 5 — HTTP server
- Zod — input validation on all tools
- TypeScript strict mode, Node.js 22, pnpm
License
MIT
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。