Headless / Coding Agent Mode
TL;DR — File a GitHub Issue describing your deployment. The Copilot Coding Agent picks it up, generates the ARM template, opens a PR, and the CI/CD pipeline handles the rest.
End-to-End Flow
How to Use It
Step 1: Create an Issue
## Deploy Request
Deploy a Python Function App with:
- Project: order-processor
- Environment: dev
- Region: eastus
- Storage Account
- Application Insights
Step 2: Agent Works Autonomously
The Copilot Coding Agent:
- Creates a branch
- Generates the ARM template at
.azure/deployments/order-processor-dev/template.json - Generates
architecture.mdwith Mermaid diagram - Runs naming validation and security checks
- Opens a PR
Step 3: CI/CD Takes Over
git-ape-plan.ymlvalidates the template and posts a what-if analysis as a PR comment- You review the plan, approve the PR
- On merge,
git-ape-deploy.ymldeploys to Azure
Step 4: Early Deploy (Optional)
Comment /deploy on an approved PR to deploy without merging first.
When to Use Headless Mode
| Scenario | Why Headless |
|---|---|
| Batch requests | File 10 issues → 10 PRs generated in parallel |
| After-hours provisioning | File issue at EOD → PR ready for morning review |
| Non-technical requestors | Business users file issues in plain English |
| Standardized workflows | Every request follows the same review pipeline |
Requirements
- Repository onboarded with Git-Ape (OIDC, environments, secrets configured)
- Copilot Coding Agent enabled on the repository
- GitHub environment protection rules for deployment safety