Skip to main content

Git-Ape

warning

EXPERIMENTAL PROJECT: Git-Ape is in active development and is not production-ready. Use it for local development, demos, sandbox subscriptions, and learning only.

Git-Ape is a platform engineering framework built on GitHub Copilot. It provides a structured, multi-agent system for planning, validating, and deploying Azure infrastructure — with security gates, cost analysis, and CI/CD pipeline integration built in.

It is the implementation of the thesis Microsoft published in Platform Engineering for the Agentic AI Era — agents and policy replacing module catalogues as the platform team's primary deliverable. See the Vision & Manifesto for the full thinking.

What it does​

  • Gather deployment requirements through guided conversations
  • Generate ARM templates and supporting deployment artifacts
  • Run security, preflight, and cost checks before deployment
  • Deploy and validate with post-deployment health checks
  • Manage lifecycle with drift detection and teardown workflows

How it works​

Git-Ape enforces compliance at three layers — generation, plan, and runtime — so non-compliant code never reaches your subscription. See Vision & Manifesto → Three layers of enforcement.

Deployment Flow​

Execution Modes​

Git-Ape works in two modes:

  • Interactive (VS Code) — Talk to @git-ape in Copilot Chat, authenticate via az login, approve each step in real time.
  • Headless (Coding Agent) — Copilot Coding Agent picks up a GitHub Issue, generates templates on a branch, opens a PR, and CI/CD workflows handle the rest.

Quick Start​

# Install the plugin
copilot plugin marketplace add Azure/git-ape
copilot plugin install Azure/git-ape

# Check prerequisites
# In Copilot Chat: /prereq-check

# Deploy something
# In Copilot Chat: @git-ape deploy a Python function app

Next Steps​

Who Is This For?​

Deep Dives​