Your AI agent writes code 10x faster. Without governance, you ship technical debt 10x faster too.
Agentic Engineering is an open-source operating system for building software with AI agents. It provides governance, project management, and engineering standards that turn AI code generation from a liability into a superpower.
AI coding agents generate code at speeds no human can match. But speed without structure creates chaos: broken conventions, missing tests, security gaps, and codebases that no one — human or AI — can maintain.
Agentic Engineering solves this with three separated concerns:
| Concern | Question | Repo |
|---|---|---|
| Governance | What may the agent do? | ai-governance-framework |
| Project Management | What should the agent do? | ai-project-management |
| Engineering Standards | How should the agent do it? | ai-engineering-standards |
A fourth repo, ai-project-templates, scaffolds all three into any new project with a single command. This repo is the umbrella: documentation, architecture maps, and the philosophy that ties the ecosystem together.
# 1. Clone the scaffolder
git clone https://github.com/EduardPetraeus/ai-project-templates.git
cd ai-project-templates
pip install pyyaml
# 2. Scaffold a new project
python scaffold.py --name my-project --stack python-data --mode solo --output-dir ../
# 3. Start working with your AI agent
cd ../my-project
claude # or open in Cursor, Windsurf, etc.Your AI agent reads CLAUDE.md automatically and operates within the defined governance boundaries.
For the full walkthrough, see the documentation site.
| Repo | Purpose | Version |
|---|---|---|
| agentic-engineering | Umbrella docs, philosophy, architecture | v1.0.0 |
| ai-governance-framework | Constitutional governance for AI agents | v0.3.0 |
| ai-project-management | YAML-based task engine for AI workflows | v1.0.0 |
| ai-engineering-standards | Code conventions, testing, git workflow | v1.0.0 |
| ai-project-templates | Project scaffolder wiring all three | v1.0.0 |
- Separate files over monolithic configuration — One file per concern. Small files are readable, diffable, and replaceable.
- Separation of concerns — Governance, project management, and engineering standards evolve at different speeds and serve different purposes.
- Templates are generic, instances are specific — The framework provides defaults. Your project customizes them.
- Machine-readable over human-only — YAML for structured data. If the agent cannot parse it, it cannot follow it.
- Solo-first, team-ready — Everything works for a single developer. The same structure scales to teams.
The full documentation site is available at eduardpetraeus.github.io/agentic-engineering.
Key pages:
- Tutorial — Progressive guide from quickstart to advanced topics
- Philosophy — Why the system is designed this way
- Architecture — How the five repos connect
- Governance in Action — Concrete examples of governance preventing real mistakes
Contributions are welcome. See CONTRIBUTING.md for guidelines.
For governance changes, read ai-governance-framework's contribution guide first — governance changes have a higher review bar.
MIT License. See LICENSE.
Built by Eduard Petraeus — shipping governance for the agentic age.