Agentwright
A 12-month, lab-driven course that takes someone with zero coding experience to deploying always-on, value-generating AI systems. Free on macOS, $0 total cost using local models, and the only published agentic-engineering curriculum built and reviewed through five expert lenses on every month.
The destination is the ability to build the system that builds the system: custom agent harnesses, software factories that produce results on spec, code that survives the churn of changing AI models, agents with safe programmatic access to the world, and agents that run 24/7 while you sleep.
How to take this course
- Clone the repository onto your Mac.
git clone https://github.com/lernen-edu/agentwright.git cd agentwright - Open the directory in your AI coding tool of choice — Claude Code, Codex CLI, Gemini CLI, Opencode, or Pi Coding Agent. The Agentwright tutor will activate automatically.
- Start at the curriculum index, Month 1, Lab 1. Work top to bottom. Update
.tutor/progress.mdas you move through the course.
See Getting Started for the full setup, including the tech stack baseline and which AI tools are supported.
What makes this different
- Built on eight evidence-based learning principles. Retrieval before reception. Worked examples that fade into independence. Spaced callbacks across months. Misconceptions named and corrected. Reflection prompts at every checkpoint.
- Authored and reviewed through five expert lenses. Curriculum architect, foundations engineer, agentic systems engineer, infra/security/devops engineer, lab designer. Every month must pass all five.
- The Socratic tutor is the companion, not the author. When you open the course in an AI coding tool, the tutor activates. It will not write your lab code. It will not run your commands. It will not paste-debug your errors. It asks back, points to course sections, and gives hints with a gap. You do the work; you learn. Read how the tutor works before you start.
- Free. Local models via Ollama; free hosted tiers where useful; paid APIs labelled with their actual dollar cost. The whole course is completable on $0.
The arc
| Phase | Months | Theme |
|---|---|---|
| Foundations | 1–5 | Command line, Git, HTTP, Python, software engineering |
| The first agent | 6 | AI APIs and the hand-built agent loop |
| Pillar 3 | 7 | Extensible Software |
| Pillar 5 | 8 | Agentic Access (and security) |
| Pillar 1 | 9 | Agent Harnesses |
| Pillar 2 | 10 | Software Factories |
| Pillar 4 | 11 | Always On Agents |
| Capstone | 12 | Integrate all five pillars into a deployed system |
What you’ll have at the end
The “AFK Value Generator” — a custom agent harness, maintained by a software factory, with pluggable model providers, secured tool access, deployed always-on, cost-tracked, and proven to run unattended for 14 days.
You’ll also have the ability to build the next one yourself, from scratch.