Azure Infrastructure MCP Server
An MCP server for managing Azure infrastructure from AI assistants, supporting subscriptions, VMs, storage, networking, identity, and more through natural language commands.
README
Azure Infrastructure MCP Server
A Model Context Protocol (MCP) server for managing Azure infrastructure directly from AI assistants like Claude in Cursor, VS Code, Claude Desktop, or any MCP-compatible client.
What You Can Do
- Manage Subscriptions — List and switch between Azure subscriptions
- Inspect Accounts — View subscription details, get access tokens, clear cached credentials
- Organize Resources — List and manage resource groups, tags, locks
- Control VMs — Start, stop, restart, deallocate virtual machines and scale VMSS
- Manage Storage — List and inspect storage accounts
- App Configuration — Manage App Configuration stores and key-values
- App Service — List and manage App Service plans and web apps
- Deploy Web Apps — Create and manage Web Apps for Containers with Docker/Podman
- Container Registry — List, create, and manage ACR instances, images, tasks, and network rules
- Virtual Networks — Create and manage VNets, subnets, and peerings
- Identity & Access — Manage Azure AD users, groups, applications, and RBAC permissions
- Governance — Work with management groups, resource locks, and tags
- Audit — View activity logs and track changes
- Docker Runtime — List, inspect logs, and restart local Docker containers
- Monitoring — System metrics, service health, and infrastructure status
Quick Start
1. Prerequisites
- Python 3.10+
- Azure CLI installed and logged in (
az login) - uv package manager (recommended) — or plain pip
2. Install
Option A — Install from PyPI (recommended):
pip install azops-mcp
# or
uv pip install azops-mcp
Option B — Run with uvx (zero-install):
uvx azops-mcp
uvx downloads the package into a cached, isolated environment and runs the server — nothing is installed permanently.
Option C — From source:
git clone https://github.com/artemkozlenkov/azops-mcp.git
cd azops-mcp
./quickstart.sh
3. Configure Your AI Client
Claude Desktop — add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):
Important: Claude Desktop does not inherit your shell's
PATH. You must use the full absolute path touvx(or any other command). Find it withwhich uvx.
Using uvx (recommended — no install required):
{
"mcpServers": {
"azops-mcp": {
"command": "/Users/YOUR_USERNAME/.local/bin/uvx",
"args": ["azops-mcp"]
}
}
}
Using a pip-installed package:
{
"mcpServers": {
"azops-mcp": {
"command": "/Users/YOUR_USERNAME/.local/bin/azops-mcp"
}
}
}
Using a local clone (development):
{
"mcpServers": {
"azops-mcp": {
"command": "/Users/YOUR_USERNAME/.local/bin/uv",
"args": ["--directory", "/full/path/to/azops-mcp", "run", "python", "-m", "azops_mcp"]
}
}
}
<details> <summary>Troubleshooting: "Failed to spawn process: No such file or directory"</summary>
This error in ~/Library/Logs/Claude/mcp-server-azops-mcp.log means Claude Desktop cannot find the binary. Claude Desktop only searches system paths (/usr/local/bin, /opt/homebrew/bin, /usr/bin, /bin) — it does not search ~/.local/bin or paths from your shell profile.
Fix: Replace "command": "uvx" with the full path from which uvx (e.g. /Users/yourname/.local/bin/uvx).
</details>
Cursor — add to ~/.cursor/mcp.json:
{
"mcpServers": {
"azops-mcp": {
"command": "uvx",
"args": ["azops-mcp"]
}
}
}
Windsurf — add to ~/.codeium/windsurf/mcp_config.json:
{
"mcpServers": {
"azops-mcp": {
"command": "uvx",
"args": ["azops-mcp"]
}
}
}
VS Code (GitHub Copilot) — add to .vscode/mcp.json in your workspace:
{
"servers": {
"azops-mcp": {
"command": "uvx",
"args": ["azops-mcp"]
}
}
}
Zed — add to your Zed settings file:
{
"context_servers": {
"azops-mcp": {
"command": {
"path": "uvx",
"args": ["azops-mcp"]
}
}
}
}
Continue (VS Code / JetBrains) — add to ~/.continue/config.yaml:
mcpServers:
- name: azops-mcp
command: uvx
args:
- azops-mcp
Note: Cursor, Windsurf, VS Code, and Zed inherit your shell's
PATH, souvxusually works as-is. If not, use the full path fromwhich uvx. See the full docs for details.
To pass environment variables (e.g. Azure credentials), add an "env" key:
{
"mcpServers": {
"azops-mcp": {
"command": "uvx",
"args": ["azops-mcp"],
"env": {
"AZURE_SUBSCRIPTION_ID": "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx"
}
}
}
}
Restart your AI client after saving the configuration.
4. Start Using
User: List my Azure subscriptions
User: Show resource groups in subscription xxx-xxx-xxx
User: Start the VM "web-server" in resource group "production"
User: What VMs are running in my dev resource group?
Authentication
| Priority | Method | When |
|---|---|---|
| 1 | Service Principal | AZURE_CLIENT_ID + SECRET + TENANT_ID all set in .env |
| 2 | Azure CLI | After az login (recommended for development) |
| 3 | Managed Identity | When running in Azure |
See the Authentication docs for the full walkthrough.
Available Tools (90+)
| Category | Tools |
|---|---|
| Health | health_check |
| Subscriptions & Auth | list_subscriptions, set_subscription, auth_status, account_show, account_clear, account_get_access_token, list_locations, list_tenants |
| Management Groups | list_management_groups, get_management_group |
| RBAC | list_role_definitions, create_role_assignment, delete_role_assignment, list_role_assignments_for_principal |
| Governance | list_resource_locks, list_tags, get_activity_log |
| Resource Groups | list_resource_groups, list_resources |
| VMs & VMSS | list_vms, get_vm_status, start_vm, stop_vm, restart_vm, deallocate_vm, scale_vmss |
| Storage | list_storage_accounts, get_storage_status |
| App Configuration | appconfig_list, appconfig_show, appconfig_kv_list, appconfig_kv_show, appconfig_kv_set, appconfig_kv_delete |
| App Service | appservice_plan_list, appservice_plan_show, webapp_list, webapp_show, webapp_start, webapp_stop, webapp_restart |
| Web Apps for Containers | webapp_create_for_container, webapp_grant_cr_access, webapp_configure_vnet_integration, webapp_assign_identity, webapp_view_logs, webapp_set_container_registry_credentials, webapp_delete |
| Container Registry | acr_list_registries, acr_show_registry, acr_create_registry, acr_delete_registry, acr_update_registry, acr_get_credentials, acr_get_login_server, acr_list_repositories, acr_list_tags, acr_show_task, acr_list_tasks, acr_create_task, acr_delete_task, acr_run_task, acr_list_builds, acr_show_quotas, acr_show_usage, acr_list_network_rules, acr_update_network_rules, acr_reset_client |
| Virtual Networks | vnet_list, vnet_show, vnet_create, vnet_delete, vnet_subnet_list, vnet_subnet_show, vnet_subnet_create, vnet_subnet_delete, vnet_peering_list |
| Azure AD (Entra ID) | aad_list_users, aad_show_user, aad_create_user, aad_delete_user, aad_list_applications, aad_create_application, aad_list_groups, aad_verify_tenant, aad_reset_client |
| Docker Runtime | list_containers, get_container_logs, restart_container |
| Monitoring | get_system_metrics, check_service_health, get_infrastructure_status |
Docker
Run the MCP server as a container:
# Build the image
docker compose build
# Run the MCP server interactively (stdio transport)
docker compose run --rm mcp-server
See the Docker docs for full instructions.
Project Structure
azops-mcp/
├── src/azops_mcp/
│ ├── __main__.py # Module entry point
│ ├── server.py # MCP server — 93 tool definitions
│ ├── config.py # Configuration management
│ ├── tools/ # Azure SDK integrations (by category)
│ │ ├── _clients.py # Shared auth & Azure SDK client factories
│ │ ├── subscription.py # Subscriptions, auth, tenants, locations
│ │ ├── resource_groups.py # Resource groups, tags, locks, activity log
│ │ ├── compute.py # VMs, VMSS, resource listing
│ │ ├── networking.py # VNets, subnets, peerings
│ │ ├── authorization.py # RBAC roles & assignments
│ │ ├── management_groups.py # Management group hierarchy
│ │ ├── app_configuration.py # App Configuration stores & key-values
│ │ ├── app_service.py # App Service plans & web apps
│ │ ├── container_registry.py # Azure Container Registry (ACR)
│ │ ├── active_directory.py # Azure AD / Entra ID
│ │ ├── webapp_deployment.py # Web App for Containers deployment
│ │ ├── docker.py # Local Docker container runtime
│ │ └── monitoring.py # System metrics & health
│ └── utils/
│ └── helpers.py # HTTP client, error formatting
├── tests/ # Unit tests (per integration)
│ ├── test_subscription.py
│ ├── test_resource_groups.py
│ ├── test_compute.py
│ ├── test_networking.py
│ ├── test_authorization.py
│ ├── test_container_registry.py
│ ├── test_active_directory.py
│ ├── test_webapp_deployment.py
│ ├── test_docker.py
│ ├── test_monitoring.py
│ ├── test_health.py
│ └── test_config.py
├── docs/ # GitHub Pages documentation
├── Dockerfile # MCP server container image
├── docker-compose.yml # Docker Compose for the MCP server
├── pyproject.toml # Dependencies & metadata
├── quickstart.sh # Setup script
└── .env.example # Configuration template
Configuration
| Variable | Default | Description |
|---|---|---|
AZURE_SUBSCRIPTION_ID |
— | Default subscription |
AZURE_DEFAULT_LOCATION |
eastus |
Default region for new resources |
LOG_LEVEL |
INFO |
DEBUG, INFO, WARNING, ERROR |
RATE_LIMIT_ENABLED |
true |
Enable rate limiting |
RATE_LIMIT_REQUESTS_PER_MINUTE |
60 |
Max requests/minute |
See .env.example for the complete list.
Development
# Install with dev dependencies
uv pip install -e ".[dev]"
# Run tests
pytest
# Code quality
black src/ tests/
ruff check src/ tests/
mypy src/
# Run server manually
uv run python -m azops_mcp
Documentation
Full documentation is available at azops.softawebit.com.
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。