This repository offers a 22-chapter "skeleton" course designed to teach users how to build and operate production AI agents. It targets engineers, researchers, and power users, providing foundational architectural patterns and decision-making frameworks. The core benefit is enabling users to learn agent design and implementation effectively by collaborating with an AI partner, focusing on durable concepts rather than rapidly evolving framework specifics.

How It Works

The course is structured as a durable skeleton of load-bearing topics, patterns, and trade-offs, designed for consumption by both humans and AI agents. Users are encouraged to pair the course material with an AI partner (like Claude Code or Codex) to explore concepts, ask probing questions, and translate patterns into practical, minimal implementations using stacks suggested by their AI. This approach prioritizes understanding fundamental agentic system architecture over specific tool implementations, ensuring long-term relevance.

Quick Start & Requirements

Clone the repository and open it in your IDE. Point your AI agent (e.g., Claude Code, Codex) at the project root. Use prompts like "Give me three real-world examples of where this matters" or "Quiz me on this topic" to engage with the material. An AI partner is the primary requirement. Optionally, run ./setup.sh to clone four reference systems (OpenCode, Hermes Agent, OpenClaw, Paperclip) for grounded examples. Links to AI behavioral guides are in CLAUDE.md and AGENTS.md.

Highlighted Details

• Features a built-in agentic-system-reviewer skill that analyzes PRDs, design docs, or agent code against the course content, providing actionable feedback with severity and evidence.
• Course chapters are logically ordered, covering Foundations, Memory, Coordination, External Surface, Production Scale, Quality/Ops, Agency, and a Design Canvas.
• Includes four optional reference open-source systems (OpenCode, Hermes Agent, OpenClaw, Paperclip) for concrete, code-level examples of discussed patterns.
• CLAUDE.md and AGENTS.md files are specifically formatted for AI assistants to read and act upon.

Maintenance & Community

The project encourages community involvement, inviting users to "Join the discord channel if you want to learn and build together!" No specific details on maintainers, sponsorships, or roadmap are provided in the README.

Licensing & Compatibility

The course content is explicitly stated as "open for educational use." The four reference systems retain their original licenses. Specific license types for the course content are not detailed, implying educational and non-commercial restrictions may apply. Compatibility for commercial use or closed-source linking is not specified.

Limitations & Caveats

This is not a step-by-step tutorial or a project-based walkthrough. It is not tied to any specific AI framework (e.g., LangChain) and does not serve as an API reference manual. The focus is strictly on architectural patterns and decision-making for agentic systems.