🎯 What You'll Learn   ✅ Prerequisites   🚀 Quick Start   📚 Journeys   🧭 What Each Journey Teaches

GitHub and Azure Agentic Journeys (Prototype)

Build and deploy applications to Azure using agents and skills.

What if you could go from idea to running in Azure and stay focused on your actual product the entire time? With agents handling the infrastructure (Bicep templates, health probes, database wiring, deployment pipelines), you can spend your time on what makes your app interesting instead of what makes it run. GitHub and Azure Agentic Journeys shows what that looks like in practice.

You pick a stack, write a spec, and hand it to your AI coding assistant, which scaffolds your API, database, and data models from the plan. From there you build out a frontend, add AI capabilities like search and recommendations through Microsoft Foundry, and let an agent generate the Azure infrastructure and deploy it when you're ready to ship. Some agentic journeys are quick deploys of open-source apps while others take you from an empty folder to a full-stack app with AI features running in production!

This is designed for:

• Developers who want to build and ship to Azure using agentic AI techniques
• Teams evaluating how agents and skills fit into their development workflow

What You'll Learn

After completing these journeys, you'll know how to:

• Build an API, database, and frontend from a spec document using GitHub Copilot
• Add AI search and a chat assistant to your app with Microsoft Foundry
• Deploy to Azure Static Web Apps, Container Apps, Functions, and AKS using Bicep generated by an agent
• Recognize when an app needs Kubernetes vs. Container Apps
• Pick your own stack (Node.js, Python, .NET, or Java) and have the agent adapt
• Deploy open-source apps (n8n, Grafana, Superset) to Azure by describing what you want to an AI agent

Prerequisites

Before starting, ensure you have:

• Agentic coding tool: GitHub Copilot CLI, the GitHub Copilot app, VS Code + GitHub Copilot, or another agentic coding tool that supports agents and skills

• GitHub account with GitHub Copilot access (Free, Pro, or Enterprise) if using GitHub Copilot

• Azure subscription - Create account

• Azure CLI (az) - Install

• Azure Developer CLI (azd) 1.28.0 or later - Install

• Node.js LTS or later - Install

  # Verify installations
  node --version
  az version
  azd version
  
  # Login once and make azd reuse the Azure CLI session
  az login
  az account show
  azd config set auth.useAzCliAuth true
  

Journey-specific tools are listed on each journey page. The cross-platform installation guide provides Windows, Mac, and Linux options for every required tool.

GitHub Copilot tools

Use whichever tool fits your workflow:

Tool	Where it runs	What it does
GitHub Copilot CLI	Your terminal	Terminal-native AI coding assistant
GitHub Copilot app	Your desktop	Agent-native workspace for parallel coding tasks
GitHub Copilot	VS Code, Visual Studio, JetBrains	Agent mode, chat, inline suggestions
GitHub Copilot cloud agent	GitHub	Assign issues to agents, get PRs back

Note

The journeys show GitHub Copilot CLI commands by default. You can also use the GitHub Copilot app or another agentic coding tool. For other tools, run: "Copy or adapt this repository's .github/skills into your supported skills or instructions location, preserving their behavior and reporting anything unsupported."

---

Quick Start

Get set up once, then pick a journey.

1. Clone and enter the repo

git clone https://github.com/DanWahlin/github-azure-agentic-journeys.git
cd github-azure-agentic-journeys

2. Confirm Azure login

az login
az account show   # confirm the subscription you expect
azd config set auth.useAzCliAuth true

3. Install the Azure plugin

Open GitHub Copilot (CLI, app, or IDE). Plugin install examples below use GitHub Copilot CLI:

copilot

Inside a CLI session (type the lines without the leading > character; in the app or VS Code, run the same plugin commands in chat):

> /plugin marketplace add microsoft/azure-skills

> /plugin install azure@azure-skills

Already installed? The plugin persists across sessions. Skip these two commands if you already ran this step.

4. Always clean up when finished

azd down --force --purge

Leaving AI Search, AKS, or SQL running is what drives most of the monthly cost.

---

Agentic journeys

Each journey is self-contained — jump into whichever one fits what you want to learn:

Journey	What you'll do	Cost if left running
WeatherView	Start here. Build an accessible vanilla JavaScript forecast app from a plan and deploy it to Azure Static Web Apps.	$0 on the Free tier
Grafana	Deploy an OSS observability app to Container Apps — one container, no database.	~$10–20/month
n8n	Deploy workflow automation with PostgreSQL, health probes, and post-provision hooks.	~$25–35/month
Apache Superset	Deploy a BI platform to AKS: init containers, ConfigMaps, Kubernetes debugging.	~$200–215/month
AIMarket	Build a full-stack marketplace from a spec: API + React + AI Search + Foundry chat.	~$100–115/month
SmartTodo	Build an AI todo app: Azure Functions, Azure SQL with managed identity, SwiftUI client.	~$10–30/month

Tip: Complete a journey and run azd down --force --purge the same day. Cost estimates are “if left on for a month,” not what you pay for a single lab session.

What each journey teaches

Pipeline slice	WeatherView	Grafana	n8n	Superset	AIMarket	SmartTodo
Idea / product intent	✅				✅	✅
PLAN/spec → agentic code	✅				✅	✅
Agent-generated Bicep + azd	✅	✅	✅	✅	✅	✅
Azure Static Web Apps	✅
Container Apps		✅	✅		✅
AKS				✅
Azure Functions						✅
Microsoft Foundry / LLMs					✅	✅
Azure AI Search					✅

---

How the agentic journeys work

Each agentic journey follows the same structure:

1. Overview diagram showing what gets created in Azure
2. Done when checklist so you know you finished
3. Agentic AI — generate code and/or deployment infrastructure
4. Verify + cleanup — prove it works, then tear down

Getting help

• App-specific issues: check the troubleshooting section in each agentic journey
• Ask GitHub Copilot
• Found a bug? Open an issue
• Want to contribute? PRs welcome.

License

This project is licensed under the MIT License.