dev tutorial · 7 modules + references · ~90 min

Build on Hermes Agent

A hands-on tour of Nous Research's open-source AI agent — written for agentic developers: people who ship real things by pairing with AI, but never sat through a compilers class. No CS degree assumed. We build the mental model first, then show you exactly where to reach in to add your own skills, tools, and integrations.

Hermes Atlas dev tutorial · Python-optional · built on the MIT-licensed hermes-agent

How to read this — each module is a self-contained 10–15 minute lesson. They build on each other, so go in order the first time. Every module cites the real source files and the official docs, and ends with a prompt you can paste to your own AI to go deeper. This is a companion to the beginner-focused handbook — the handbook gets you using Hermes; this tutorial gets you building on it.

Foundations

What Hermes Agent actually is — the mental model (an agent is a loop, not a chatbot), what makes it different, and getting it running in five minutes. ~12 min
The agent loop — the heartbeat — follow one message from your keyboard to the model and back. Tool calls, the loop, when it stops. The single most important concept. ~14 min

The extension points

Tools & toolsets — what a "tool" really is, the self-registering registry, how tools get grouped, and the command-approval safety model. ~14 min
Skills — the self-improvement loop — the headline feature, and your easiest way in. Author a skill by hand (it's just Markdown), then watch the agent improve its own. ~15 min
Memory & cross-session recall — how the agent remembers you between sessions: MEMORY.md, the user profile, full-text session search, and pluggable memory backends. ~11 min

Reach & building

Reaching out: MCP, the gateway & cron — give the agent more tools (MCP), let it live on Telegram/Discord (the gateway), and have it work while you sleep (cron). ~13 min
Building on Hermes: pick your altitude — the capstone. The "Footprint Ladder" for choosing how to extend, embedding the agent in your own Python app, and delegating to subagents. ~15 min

Reference — keep these open while you build

Glossary — every Hermes term used in the tutorial, in plain English.
Codebase map & cheat sheet — the "where do I edit X?" companion: key files, your ~/.hermes/ data dir, and essential commands.
Skill authoring template — a copy-paste SKILL.md with annotated frontmatter and a finished example.

Your AI is your TA — this whole tutorial assumes you'll be pairing with an AI assistant. Clone the repo alongside these lessons and let your AI read the real files with you. Start with: "I'm learning the Hermes Agent codebase (github.com/NousResearch/hermes-agent). I cloned it locally. Walk me through agent/conversation_loop.py and explain the main loop like I'm a product builder, not a compiler engineer."

Start Module 1 · What Hermes Agent actually is →