PTENES
Skip to content
MODULE 3.2

🔭 OpenClaw, Hermes, and agentic systems

You don’t need to build your Jarvis from scratch. There are already ready-made systems — and INEMA has courses dedicated to several of them. This module is a overview: what each one is, how they differ, and how to choose yours without getting lost.

6
Topics
~50
Minutes
Advanced
Level
Theory
Type
Progress in this module0%

0 of 6 sections read

💡 New here? Three words before you start

  • Agentic system — a ready-made “Jarvis” package: brain + memory + skills already set up, which you just configure.
  • Self-hosted — “hosted by you”: it runs on your machine or server instead of on a third-party website.
  • Docker — a way to “package” a program so it runs the same on any computer without installation headaches.
1

🧱 Ready-made agentic systems: what they are

In module 3.1, you saw that a Jarvis has four pieces: a brain, memory, skills, and triggers. Setting all of that up by hand takes work. The good news: other people have already put together and made them ready for you to use. These kits are the ready-made agentic systems.

Imagine an apartment: you can buy the land and build it brick by brick, or get a furnished apartment and just adjust the decor. A ready-made agentic system is like a furnished apartment—the walls are already up; you just configure what’s yours.

🔑 The central idea

The difference between build e set up:

  • •From scratch: more control, more time, more things that can break.
  • •Ready: gets started quickly, trading a little control for speed.
📦 🧠 brain 💾 memory 🧰 skills ⚡ triggers
A ready-made system arrives with all four pieces in place. Your job is no longer to build and becomes set up.

✓ When to use a ready-made one

  • ✓You want quick results, not a project
  • ✓Your needs are common (reply, remember, schedule)
  • ✓There’s an INEMA course teaching that system

✗ When to build from scratch

  • ✗You need something very specific or unusual
  • ✗Total control matters more than speed
  • ✗You want to learn how each piece works inside

Key concepts

📌 Ready-made kit = furnished apartment
📌 You sets up, don't build
📌 Speed vs. control
📌 It can save months
2

🪶 Hermes (Nous Research): open AI for building assistants

O Hermes is a family of AI models open, created by Nous Research. "Open" means it's public and you can run your own copy — including on your own server (self-hosted). It's one of the most widely used foundations for anyone who wants a Jarvis that doesn't depend on just one company.

Think of Hermes as the engine of the car. It isn’t the whole car — it’s the part that thinks and responds. Around it, you build the memory, skills, and triggers. It’s a great path for anyone who values privacy and control.

🎓 Explore these INEMA courses in more depth

This module is just an overview. To actually build an assistant with Hermes, INEMA has two dedicated courses:

  • •Hermes 21C — the introductory course on the Hermes model and how to get started.
  • •Hermes Agent — how to turn Hermes into an agent that executes tasks.

💡 Practical tip

If “privacy” and “running on my machine” matter to you, Hermes is usually the place to start. Start with the Hermes 21C before the Hermes Agent—one is the foundation for the other.

Key concepts

📌 AI open (Nous Research)
📌 Runs self-hosted
📌 It’s the "engine," not the whole car
📌 Strong on privacy/control
3

🐾 OpenClaw: autonomous multi-channel assistant

O OpenClaw it’s an AI assistant autonomous e multi-channel. Autonomous: it doesn’t just wait for instructions — it takes initiative to complete a task. Multichannel: you can talk to it in several places (messages, email, etc.), not just in one little window. It usually runs via Docker, that way of packaging everything so it's easy to install.

If Hermes is the engine, OpenClaw is closer to the assembled car: already comes with the “serve multiple channels” part handled. It’s a good option when you want a Jarvis that talks to you wherever you already are.

🎓 Explore this INEMA course in more depth

Here, this is just an overview. To get OpenClaw running in practice, with Docker and everything, follow:

  • •Docker OpenClaw — the INEMA course that teaches you how to deploy OpenClaw via Docker, step by step.
🪶 Hermes open · self-hosted thinking engine 🐾 OpenClaw autonomous · multi-channel via Docker 🗺️ Others several projects broad overview
Three useful families side by side: Hermes (the engine), OpenClaw (the ready-made multi-channel assistant), and the rest of the landscape. None is “the best” — it depends on what you need.

💡 Practical tip

"Docker" seems intimidating from a distance, but the course Docker OpenClaw treats you as a beginner: it’s more “copy, paste, and run” than programming. If your goal is to talk to Jarvis on your phone, this is the way to go.

Key concepts

📌 Autonomous (takes initiative)
📌 Multi-channel (says where you are)
📌 Runs via Docker
📌 More "assembled car" than engine
4

🗺️ Other paths: the landscape

Hermes and OpenClaw are two paths with an INEMA course, but the landscape is broader. There are other ways to have a Jarvis—and the goal here isn’t about memorizing names, but for you to recognize the families and not be intimidated when you hear about them.

Several connected AI cores, representing the ecosystem of ready-to-use agentic systems for building a Jarvis.
The agentic systems ecosystem: many paths lead to a Jarvis. The fun is choosing one, not collecting them all.
A

🪶 Open models (like Hermes)

Public AI that you run yourself. More control and privacy, but it takes a little more hands-on work.

B

🐾 Ready-made assistants (like OpenClaw)

Kits that already handle channels and automation. They get you started quickly, trading some control for convenience.

C

☁️ Cloud platforms

Services that host Jarvis for you. Easiest to get started with; less privacy, recurring cost.

D

🧩 Build from scratch SEEN IN 3.1

All four pieces at your fingertips. Maximum control, more work — good for those who want to learn how it works under the hood.

🎯 The pattern to watch for

All these paths are the same four pieces put together in different ways. Anyone who understood module 3.1 can understand any new system: just ask, “where are the brain, memory, skills, and triggers?”

Key concepts

📌 Recognize families, not names
📌 Open · ready-made · cloud · from scratch
📌 They’re the same 4 building blocks
📌 Map before choosing
5

⚖️ How to choose yours

There’s no “best system”—there’s the best for you. The choice comes down to four simple questions about what you need: privacy, channels, ease of use, and cost. Answer those four, and the path practically chooses itself.

Watch out for a trap: the temptation to want “the most powerful.” The most powerful is usually the most work. To get started, ease beats power almost always.

✓ Good selection criteria

  • ✓"Do my data need to stay with me?" → Hermes/self-hosted weighs more heavily
  • ✓"Do I want to talk through multiple channels?" → OpenClaw weighs more heavily
  • ✓"Is there an INEMA course that teaches this?" → you can go further on your own
  • ✓"How much can I spend per month?" → rules out expensive options

✗ Misleading criteria

  • ✗"Which one is the most powerful?" — power ≠ useful to you
  • ✗"Which one is most trendy?" — trends pass; your needs remain
  • ✗"Which one has the most features?" — a feature you don’t use is dead weight

💡 Practical tip

Write it in one line: "I want a Jarvis that [does what], where [privacy or convenience] matters more, talking to me through [channel], spending up to [amount]." This single sentence already points to the right system.

Key concepts

📌 There is no "best"; there is "yours"
📌 4 questions: privacy, channels, ease of use, cost
📌 Ease > power (at first)
📌 INEMA course = go further
6

🚪 Where to start without getting lost

The biggest trap at this stage isn't technical: it's the option paralysis. There are so many ways that you feel like trying them all—and then you finish none. The antidote is simple: choose ONE and complete the corresponding INEMA course.

1️⃣
Define your statement of need (the one from topic 5). It points to the system.
2️⃣
Choose ONE path — privacy/control → Hermes; ready-made multi-channel setup → OpenClaw.
3️⃣
Follow the INEMA course for that path from start to finish, without skipping to another one.
4️⃣
Only afterward once you have something running, consider trying another.

🎓 Your next steps (INEMA courses)

✋ Before you continue — what's the best way to get started with agentic systems?

Key concepts

📌 Enemy = option paralysis
📌 Choosing one and follow
📌 INEMA course from start to finish
📌 Focus beats collecting tools

🎓 Module summary

✓
Ready-made Jarvis assistants exist — you configure it instead of building it from scratch.
✓
Hermes — open AI (Nous Research), self-hosted; Hermes 21C and Agente Hermes courses.
✓
OpenClaw — a multi-channel autonomous assistant via Docker; Docker OpenClaw course.
✓
Choose based on your needs — and start with ONE path, without getting lost.

Next module:

3.3 — 🧰 Ready-made skills: a library to copy, paste, and adapt for your Jarvis.