az prototype

Rapid Azure prototype generation powered by AI agent teams.

Overview

az prototype is an Azure CLI extension that empowers customers to rapidly create functional Azure prototypes using AI-driven agent teams. It supports three AI providers — GitHub Copilot, GitHub Models, and Azure OpenAI — for intelligent code and infrastructure generation.

Prerequisites

• Azure CLI 2.50+
• Azure subscription with appropriate permissions
• GitHub CLI (gh) installed and authenticated (required for copilot and github-models providers)
• GitHub Copilot license, Business or Enterprise (required for copilot provider only)

Installation

NOTE: Currently, AZ Prototype is in preview. We are aggressively working to produce our first stable version in March 2026. Please log all bugs in Issues, and we will address them as soon as possible.

Install

az extension add --name prototype

Upgrade

Only Stable Versions:

az extension update --name prototype

Include Preview Versions:

az extension update --name prototype --allow-preview

Quick Start

# Initialize a new prototype project
az prototype init --name my-prototype --location eastus

# Run design analysis (interactive)
az prototype design

# Provide artifacts for design analysis
az prototype design --artifacts ./requirements/ --context "Build a data pipeline"

# Generate infrastructure and application code
az prototype build

# Deploy to Azure (incremental — only deploys changes)
az prototype deploy

# Build only infrastructure code
az prototype build --scope infra

# Build only application code
az prototype build --scope apps

To receive help for any specific command, run az prototype --help or az prototype <command> --help.

View the command reference to see the full list of commands and their parameters.

Agent System

Built-in Agents

Ships with 19 pre-defined agents:

Agent	Capability	Description
cloud-architect	Architecture	Overall overseer, Core layer ownership, deployment planning
infrastructure-architect	Infrastructure	Infrastructure layer ownership, directs terraform/bicep agents
data-architect	Data	Data layer ownership, schema and access patterns
application-architect	Application	Application layer ownership, developer delegation
security-architect	Security	Cross-cutting security, RBAC, identity, encryption
terraform-agent	Terraform	Terraform IaC generation (azapi provider)
bicep-agent	Bicep	Bicep template generation
csharp-developer	C# / .NET	C# application code generation
python-developer	Python	Python application code generation
react-developer	React / TypeScript	React frontend code generation
app-developer	Development	Generic fallback for unsupported languages
doc-agent	Documentation	Project and deployment documentation
qa-engineer	QA / Analysis	Error diagnosis, code review, remediation
biz-analyst	Business Analysis	Requirements gap analysis and interactive design dialogue
cost-analyst	Cost Analysis	Azure cost estimation at S/M/L t-shirt sizes
project-manager	Coordination	Scope management, task assignment, escalation
monitoring-agent	Monitoring	Observability configuration generation
governor	Governance	Embedding-based policy retrieval and enforcement
advisor	Advisory	Per-stage trade-off and risk analysis

Custom Agents

Add your own agents via YAML or Python:

# List available agents
az prototype agent list

# Add a custom agent from YAML
az prototype agent add --file ./my-agent.yaml

# Override a built-in agent
az prototype agent override --name cloud-architect --file ./my-architect.yaml

YAML Agent Format

name: my-custom-agent
description: Custom agent for specific use case
role: architect
system_prompt: |
  You are a specialized architect for ...
constraints:
  - Must use managed identity
  - Must follow naming conventions
tools:
  - terraform
  - bicep

Configuration

Project configuration is stored in prototype.yaml:

project:
  name: my-prototype
  location: eastus
  environment: dev
  iac_tool: terraform  # or bicep

naming:
  strategy: microsoft-alz  # microsoft-alz | microsoft-caf | simple | enterprise | custom
  org: contoso
  env: dev
  zone_id: zd              # ALZ zone ID (see table below)

ai:
  provider: copilot  # copilot | github-models | azure-openai
  model: claude-sonnet-4

agents:
  custom_dir: ./.prototype/agents/
  overrides: {}

deploy:
  track_changes: true

Naming Strategies

All agents use a shared naming resolver to generate consistent Azure resource names.

Strategy	Pattern	Example
microsoft-alz (default)	{zoneid}-{type}-{service}-{env}-{region}	zd-rg-api-dev-eus
microsoft-caf	{type}-{org}-{service}-{env}-{region}-{instance}	rg-contoso-api-dev-eus-001
simple	{org}-{service}-{type}-{env}	contoso-api-rg-dev
enterprise	{type}-{bu}-{org}-{service}-{env}-{region}-{instance}	rg-it-contoso-api-dev-eus-001
custom	User-defined	Depends on pattern

Landing Zone IDs (ALZ)

When using microsoft-alz, resources are assigned to a landing zone:

Zone	ID	Used for
Connectivity Platform	pc	Networking, DNS, firewall
Identity Platform	pi	Entra ID, RBAC
Management Platform	pm	Log Analytics, App Insights
Development Zone	zd	Dev workloads (default)
Testing Zone	zt	QA / test workloads
Staging Zone	zs	UAT / staging workloads
Production Zone	zp	Production workloads

Stages

Command	Description	Re-entrant
az prototype init	Project scaffolding, auth, config	No
az prototype design	Requirements analysis, architecture	Yes
az prototype build	Generate IaC and app code	Yes
az prototype deploy	Deploy with change tracking	Yes

Development

The Azure CLI prototype extension is architected and developed by Joshua Davis, and it is based on a solution engineering program he launched within Microsoft enterprise field sales, the Innovation Factory, that delivers rapid prototypes for Microsoft enterprise customers.