Azure Agentic Infraops Accelerator
Starter template for APEX. Contains custom Copilot agents, Azure Pricing MCP server, and an orchestrated multi-step workflow. Auto-syncs from the main repository.
Ask AI about Azure Agentic Infraops Accelerator
Powered by Claude Β· Grounded in docs
I know everything about Azure Agentic Infraops Accelerator. Ask me about installation, configuration, usage, or troubleshooting.
0/500
Reviews
Documentation
APEX Accelerator
Modernize your Azure Infrastructure with AI. A production-ready template for building Well-Architected environments using custom Copilot agents, Dev Containers, and the Model Context Protocol (MCP).
Overview
This accelerator provides the scaffolding and governance to move from requirements to deployed infrastructure using an orchestrated multi-agent workflow. It leverages domain-specific AI agents to ensure every deployment is Well-Architected, governed, and documented.
What you get: Specialized agents, skills, validation scripts, a full dev container with all tools pre-installed, and an optional weekly sync workflow that keeps your agents and skills up to date with the upstream APEX project.
Prerequisites
| Requirement | Details |
|---|---|
| VS Code | Latest stable release |
| GitHub Copilot | Active license (Individual, Business, or Enterprise) |
| Docker Desktop | For the dev container (or GitHub Codespaces) |
| Azure subscription | Optional for Steps 1-5; required for Step 6 (Deploy) |
Quick Start
1. Create Your Repository
This repository is a GitHub Template β not a fork.
- Click "Use this template" β "Create a new repository" at the top of this page
- Choose an owner and name (e.g.,
my-infraops-project) - Select Private (do not use Public)
- Click Create repository
Your new repo has the same directory structure and files but a clean commit history and no fork relationship. It is entirely yours.
2. Clone and Open in Dev Container
git clone https://github.com/YOUR-USERNAME/my-infraops-project.git
cd my-infraops-project
code .
When prompted by VS Code, click "Reopen in Container" (or run Dev Containers: Reopen in Container
from the Command Palette). The container build takes 3-5 minutes and pre-installs:
- Azure CLI with Bicep extension
- Terraform CLI with TFLint
- GitHub CLI (
gh) - Node.js + npm (validation scripts)
- Python 3 + pip (MCP server, diagram generation)
- Go (Terraform MCP server)
3. Initialize Your Repository
After the dev container starts, run the initialization commands:
npm install
npm run init
npm run sync:workflows
What these do:
| Command | Purpose |
|---|---|
npm install | Install Node.js dependencies (validation scripts, linting) |
npm run init | One-time setup β replaces all references to the accelerator template repo with your repo's URL (auto-detected from git remote). Run npm run init -- --dry-run first to preview changes. |
npm run sync:workflows | Fetches the latest GitHub Actions workflows from the upstream APEX project into your .github/workflows/ directory |
Note: Python dependencies (diagrams, Azure Pricing MCP server, apex-recall) are installed automatically by the dev container's
post-create.shscript. No manualpip installis needed.
After running, review and commit:
git diff
git add -A && git commit -m "chore: initialize from template"
4. Set Up Azure (Optional)
Run the setup wizard to configure Azure OIDC authentication, RBAC roles, and GitHub secrets/variables β all in one command:
az login
npm run setup
The wizard creates an Entra ID app registration, OIDC federated credentials (for main branch + dev/staging/prod environments), assigns Reader at your Management Group and Contributor at your subscription, and configures all GitHub secrets and variables. It is idempotent β safe to re-run.
See the Azure Setup documentation for headless mode, manual setup steps, and troubleshooting.
Skip this if you are just exploring the agent workflow without deploying to Azure.
Project Structure
Once you are working in your repo, here is what lives where:
.github/
agents/ # Agent definitions (*.agent.md)
_subagents/ # Subagent definitions (non-user-invocable)
skills/ # Reusable domain knowledge (SKILL.md per skill)
instructions/ # File-type rules with glob-based auto-application
copilot-instructions.md # VS Code Copilot-specific orchestration instructions
workflows/ # GitHub Actions (sync, CI, validation)
agent-output/ # Generated artifacts organized by project
{project}/ # 01-requirements.md through 07-*.md
infra/
bicep/{project}/ # Bicep templates (main.bicep + modules/)
terraform/{project}/ # Terraform configurations (main.tf + modules/)
tools/
mcp-servers/ # MCP servers (azure-pricing, drawio)
scripts/ # Validation and maintenance scripts
site/ # Documentation site source (Astro Starlight)
What's Yours vs. What's Upstream
| Your files (safe to edit, never overwritten by sync) | Upstream-managed files (overwritten by sync) |
|---|---|
agent-output/ β generated project artifacts | .github/agents/ β agent definitions |
infra/bicep/ β your Bicep templates | .github/skills/ β agent skills |
.github/workflows/ β your CI/CD workflows | .github/instructions/ β coding rules |
README.md β your documentation | .github/copilot-instructions.md |
tools/, site/, package.json | |
AGENTS.md, .devcontainer/, .vscode/ |
Note: If you disable the sync workflow, everything becomes yours to edit freely. See Customization below.
Customization
Changing defaults (regions, tags, naming)
When you create from this template, every file is yours. The question is which approach gives you the best experience over time.
Strategy A: Edit directly (simplest)
Edit the file you need β for example, change swedencentral to westeurope in
.github/skills/azure-defaults/SKILL.md. If you don't plan to pull upstream
improvements, disable the sync workflow entirely (repo Settings β Actions β disable
Upstream Sync) and manage the repo as your own.
If you want upstream updates later, re-enable the workflow. The sync PR will overwrite your edited files with the upstream version (it is not a merge), so you'll need to re-apply your changes after merging. Keep a record of what you changed.
Strategy B: Keep sync enabled, layer your overrides (recommended for teams that want continuous upstream improvements)
The sync workflow overwrites all upstream-managed files but never touches the four excluded paths. To safely store overrides that survive sync:
-
Edit the sync exclusion list β the sync workflow file itself is user-owned (
.github/workflows/is excluded from sync). Open.github/workflows/weekly-upstream-sync.ymland add paths toEXCLUDE_PATHSand the matchingfor path in ...loop:EXCLUDE_PATHS: | .github/workflows/ agent-output/ infra/bicep/ infra/terraform/ README.md AGENTS.mdThen add your overrides to root
AGENTS.mdβ it will survive all future syncs. -
Use
infra/bicep/AGENTS.mdβ sinceinfra/bicep/is already excluded from sync, you can place anAGENTS.mdthere with your organization defaults. VS Code loads subfolderAGENTS.mdfiles when working with files in that directory. -
VS Code user-profile instructions β place a
.instructions.mdfile in your VS Code profile'sprompts/folder. This lives outside the repo entirely and applies to all your workspaces.
Which strategy to pick?
| Approach | Best for |
|---|---|
| A: Edit directly, skip sync | Solo users, teams that self-manage updates |
| B: Layer overrides, keep sync | Teams that want automatic upstream improvements |
Adding new agents or skills
Agents are defined in .github/agents/*.agent.md and skills in
.github/skills/*/SKILL.md. You can add new ones alongside the existing set.
If you keep sync enabled, use distinctive names that won't collide with upstream
filenames, or add your custom paths to the sync exclusion list.
IaC Tracks: Bicep and Terraform
Both IaC tracks are fully supported. The Requirements agent (Step 1) captures your
iac_tool preference, and the Orchestrator routes Steps 4-6 to the correct track.
| Factor | Bicep | Terraform |
|---|---|---|
| Azure-only | Native DSL, first-class support | Multi-cloud via AzureRM provider |
| State | No state file (ARM-managed) | State file (Azure Storage backend) |
| AVM modules | br/public:avm/res/ | registry.terraform.io/Azure/avm-res-*/ |
| CI/CD | az deployment group create | terraform plan + terraform apply |
Multi-Project Support
The accelerator is designed for one repo containing multiple projects. Each project gets its own folders:
agent-output/{project}/β artifacts (requirements, architecture, plans, docs)infra/bicep/{project}/orinfra/terraform/{project}/β IaC templates
Agents, skills, instructions, and the dev container are shared across all projects.
One repo per project is also valid when teams need separate governance or isolation. Each repo is created independently from the template.
Sharing customizations across repos or teams
- Edit the sync exclusion list in each repo to protect
AGENTS.md, then maintain a standard overrides section that you copy into each repo - VS Code user-profile instructions β personal preferences that follow you across repos without any per-repo setup
- Canonical overrides in a shared location β maintain a standard overrides snippet in a team wiki or internal repo and copy it when creating new instances
Keeping Up to Date
| What | How | Frequency |
|---|---|---|
| Agents, skills, instructions, docs, scripts | Automated weekly PR (upstream sync workflow) | Weekly |
| GitHub Actions workflows | npm run sync:workflows (manual) | As needed |
| All validations | npm run validate:all | Before each PR |
The sync workflow opens a PR for human review β it never auto-merges. You can disable it entirely if you prefer to manage updates manually.
Validation & Quality
# Run all code and documentation validations
npm run validate:all
# Fix markdown formatting
npm run lint:md:fix
# Bicep validation (replace {project})
bicep build infra/bicep/{project}/main.bicep
bicep lint infra/bicep/{project}/main.bicep
# Terraform validation
terraform fmt -check -recursive infra/terraform/
cd infra/terraform/{project} && terraform init -backend=false && terraform validate
Resources
- APEX upstream project
- APEX documentation
- Azure Setup guide
- MicroHack (hands-on exercises)
- Prompt Guide
- FAQ
