az prototype

An Azure CLI extension that enables users to rapidly build Azure proof-of-concept deployments using AI-driven agent teams. It implements a condensed version of the Innovation Factory POC methodology that Microsoft sales uses internally — 4 stages, 20 agents, production-quality output.

Why use it?

• Go from idea to deployed Azure prototype in minutes, not days
• AI agents handle architecture, IaC generation, deployment, and troubleshooting
• Governance policies, anti-patterns, and standards enforce best practices automatically
• Every stage is re-entrant — iterate and refine without starting over

---

Pipeline

The CLI condenses the Innovation Factory's 12 stages into 4 core commands:

init  ──>  design  ──>  build  ──>  deploy

Stage	What it does
init	Create project folder, scaffold configuration (prototype.yaml)
design	Interactive discovery conversation, requirements analysis, architecture design, deployment planning
build	Generate IaC (Bicep or Terraform) and application code from the design
deploy	Run preflight checks, deploy infrastructure, capture outputs, run verification

Each stage is re-entrant. Re-run design after deployment to refine architecture based on feedback, or re-run build to regenerate code after design changes.

---

Features

• 20 AI agents — 5 architects (cloud, infrastructure, data, application, security), 3 language-specific developers (C#, Python, React) + generic fallback, 2 IaC agents (Terraform, Bicep), and 9 supporting agents (QA, cost, docs, monitoring, governance, advisory, biz-analyst, project-manager, security-reviewer)
• 5 workload templates — web-app, serverless-api, microservices, ai-app, data-pipeline
• Governance engine — 58 policy rules, 40 anti-pattern checks, 38 design standards enforced during generation
• Benchmark suite — 14 quality benchmarks (B-INST through B-ANTI) for measuring AI-generated code quality, with HTML dashboard, PDF reporting, and trend tracking
• TUI dashboard — Rich interactive terminal UI for design, build, and deploy sessions
• Cost analysis — S/M/L tier estimation via the cost-analyst agent
• Backlog generation — Generate and push user stories to GitHub Issues or Azure DevOps
• Four-level taxonomy — Layer/Capability/Component/Resource hierarchy drives deployment ordering and agent ownership (Layer Architecture, Application Architecture)
• MCP integration — Model Context Protocol plugin system for extending agent capabilities
• Knowledge system — Runtime documentation, web search, and self-learning contributions
• Error analysis — QA-first troubleshooting with automatic escalation
• Docs and spec kit — Generate project documentation and stakeholder-ready specification packages

---

Quick Start

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

# 2. Run interactive design session (discovery + architecture)
az prototype design

# 3. Generate infrastructure-as-code and application code
az prototype build

# 4. Deploy to Azure
az prototype deploy

See Installation for setup instructions and Quickstart for a full walkthrough.

---

Command Reference (Summary)

Command Group	Commands
az prototype	init, launch, design, build, deploy, status
az prototype analyze	error, costs
az prototype config	init, show, get, set
az prototype generate	backlog, docs, speckit
az prototype knowledge	contribute
az prototype agent	list, add, override, show, remove, update, test, export

See Command Reference for full details on parameters and usage.

---

Innovation Factory Stage Mapping

This CLI condenses the Innovation Factory's 12 detailed stages into 4 re-entrant stages:

CLI Stage	IF Stages	Purpose
init	--	Project folder initialization, config scaffolding
design	1-6	Discovery conversation, requirements analysis, architecture design, deployment planning
build	7	Generate IaC (Bicep/Terraform) and application code
deploy	8-10	Infrastructure deployment, app deployment, customer testing
design (re-run)	10-11	Refinements based on feedback, architecture improvements

---

Navigation

Section	Description
Installation	Prerequisites, extension install, AI provider setup
Quickstart	End-to-end walkthrough from init to deploy
Stages	Detailed documentation for each pipeline stage
Configuration	prototype.yaml, secrets, AI providers, naming strategies
Agent System	Built-in agents, custom agents, overrides, governance
Layer Architecture	Four-level taxonomy, deployment ordering, layer ownership
Application Architecture	App sub-layers, developer delegation, project structure
Templates	Workload templates and customization
Backlog Generation	Generate and push stories to GitHub/Azure DevOps
MCP Integration	Extend agents with Model Context Protocol plugins
Troubleshooting	Common issues and solutions