Welcome

Welcome to the most comprehensiveHermes Agentcourse available. Hermes Agent is an AI 'employee' that runs on your own computer, connects to your channels, remembers your preferences over time, and does real work for you — autonomously and 24/7.

This course teaches you everything needed to build andsellcomplete AI-agent systems for clients, charging from$500 to $5,000 per setupand$200 to $1,000 per monthfor maintenance.

What You Will Learn

• What Hermes Agent actually is and how it differs from other agents (OpenClaw, Claude Code, Codex)
• The good, the bad, and the ugly of autonomous agents
• Full setup from scratch — hosting, costs, and decisions
• The 'Buddy System' using Codex as your repair doctor
• Connecting channels: Telegram, Slack, iMessage, Discord
• Identity files, personality (soul.md), and user memory
• Shared memory with Honcho
• Managing multiple agents with Multica (project boards)
• Connectors: MCPs, CLIs, and APIs (Composio, Google Workspace CLI)
• Cron jobs, skills, and agentic workflows (autopilots)
• Practical builds: morning briefings, animated websites, property marketing kits, carousels, content repurposing, e-com ads, community managers, trading bots, health coaches, and Jarvis command center
• How to price, package, and sell these systems

How to Use This Course

1. Watch in orderif you're brand new — the lessons build on each other.
2. Build alongside— the single most effective way to learn is to follow along and create your own Hermes setup.
3. If you already have Hermes, jump to the specific section you need using the timestamps from the source video.

The Big Opportunity

"The gap between people who use these agents and people who don't is getting a lot wider. Every company is going to need something like this, and most have no idea it barely even exists."

There are 33 million companies in the U.S., and roughly 30 million are solopreneurs — single-person enterprises that won't have AI budgets or AI experts. The people who learn how to implement AI agents now are the ones who will build the businesses and careers everyone wishes they started before.

Next up:What Hermes Agent actually is.

The High-Level Idea

At a high level: take the smartest AI model available (ChatGPT, Claude, Grok, etc.), give iteyes and hands, and sit it at a desk with a mouse connected to your computer. Now you can message it like a coworker, and it can do everything a regular person could do sitting at that computer — on its own, without you asking it constantly.

Hermes Agent vs. Other Agents

Hermes is similar toOpenClaw, but it is built on fundamentally different philosophy:

• OpenClaw: "Here's an AI agent that can use your computer."
• Hermes Agent: "Here's an AI employee that grows with you over time."

The Self-Improvement Loop

The biggest difference is Hermes' automaticself-improvement loop. If Hermes learns how to do something for you — like how you edit your videos or how to write in your tone — itautomatically saves that process as a reusable skill. It does this on its own, without you manually asking.

This means:1. You give Hermes a task.2. It does the task.3. It figures out a better way, or you correct it.4. It remembers that as askill.

Example: While apartment-hunting in New York, Hermes noticed a preference for high-rises within a certain budget. Later, when searching again, it automatically suggested similar listings — without being asked.

Memory Curation

Every ~10 turns or so, Hermes runs a process in the background that improves its own memory. It has a 'curator' that:- Cleans up old skills- Improves existing skills- Archives things that are no longer useful

This keeps the agent from getting messy and bloated over time — a problem OpenClaw used to have. Hermes doesn't just blindly remember everything; itstrategically archives, staying effective at the things you work on day-to-day.

Why People Switch

It's not because Hermes has one flashy feature. It's because itfeels more like a real employeeand gets more useful the longer you use it.

Next up:Model vs. Agentic Harness — the car analogy.

Two Different Things

Before installing anything, it's worth understanding the difference between amodeland anagentic harness. The easiest way to think about it is like acar:

• The model is the engine.It's powerful, but by itself it just sits there — it has no wheels, steering, or brakes to actually go anywhere.
• Hermes is the rest of the car.It wraps around the model and lets it actually do the work.

What a 'Model' Means

When we saymodel, we mean the AI intelligence layer:- GPT 5.5- Claude Opus- Grok- Google Gemini

The model provides the intelligence, but not the ability to act on the computer.

What Hermes (the Harness) Does

Hermes is theagentic harnessthat comes bundled with:- The ability to control everything on your computer- Runs on the machine 24/7- Gateways to start messaging it on many different channels- The ability to get work done like an employee- Its own memory system

Why Not Just Claude Code or Codex?

There are other good harnesses (Claude Code, Codex), but they were designed for different jobs:

• Claude Code— as the name suggests, was built tocode.
• Hermes Agent— was designed to be anAI employee, built to know you, your business, your customers, your preferences, and your workflows over time, thanks to how its memory system was built.

Big Advantages of Hermes

1. Model agnostic— you are not locked into one provider. As models leapfrog each other, you always use the best model for each task.
2. Local— all your data lives on your computer, right next to you.

Next up:The Good, the Bad & the Ugly.

You Are Still in Charge

There's a version where it sounds like you hand Hermes the keys and disappear — but that's not how any of this works. These agents are powerful butnot autonomous decision-makers. You can't let them loose without structure.

The Bad & the Ugly

It's on you to:- Give the agent theright access.- Protect yourself from giving ittoo much, too fast.

Real warning:A researcher gave her agent access to her email inbox, and the agent deleted it. She lost years of emails. This actually happened.

Key Rule: Verify Outputs

Alwaysverify outputs before you trust them. Put safeguards in place before you let your agent:- Send an email- Move any money- Publish anything

The course shows how to set up these safeguards.

The Opportunity Gap

The reality right now is that the gap between people who use agents and people who don't is widening fast.

• Every company will need something like this.
• Most companies have no idea it barely exists or how to start.
• People who understand this now, while it's still early, are the ones who end up building businesses and careers.

What People Are Already Selling

The same systems built in this course are being sold to businesses as'done-for-you' digital employees:

• $500–$5,000 per monthto set it up
• $200–$1,000 per monthto maintain it

The business owner doesn't need to understand how it works — they just get an agent that runs 24/7 at a fraction of what a human would cost.

Next up:Full setup from scratch — hosting & cost decisions.

Two Hosting Paths

You can run your Hermes Agent in two ways:

1. VPS (cloud server)— host a server in the cloud.
2. A personal computer— a Mac Mini, old computer, or PC (Mac or Windows both work).

Recommendation: Use an Old Computer

The course recommends a personal computer (like a Mac Mini) because:- It gives you an easy interface- It's easier to control- You can see what's happening on your screen- You pay for the hardwareonly once

Note on cons:The power needs to stay on, and access is limited offline. The 'Buddy System' (Codex) solves this — more on that next.

Choosing a 'Brain' (Model Provider)

After choosing the hardware, choose abrainfor your Hermes. A common recommendation is aChatGPT / OpenAI subscription, which gives the most value for intelligence, accessibility, and ability.

Managing Cost: Two Billing Models

1. Subscription (Recommended for Beginners)

Using aChatGPT subscriptiongets you unlimited access within a time window. Benefits:- More usage-Flat costs(predictable)- Protects you from racking up surprise bills

2. API / Pay-Per-Usage

This is when someone asks you for anAPI keyto run Hermes. If you don't know what you're doing and something runs that you don't want, it can rack uphundreds or thousands of dollarsin credits. This is a real risk for beginners.

The course shows how to use the ChatGPT subscription to get Hermes running affordably — you can even start free and scale up as needed.

Next: The Buddy System

Before installing, set up theBuddy Systemusing Codex as your repair doctor.

Next up:The Buddy System: Codex as your repair doctor.

What Is the Buddy System?

You'll use Hermes andCodexwith the same brain (GPT 5.5 via your subscription). Codex becomes your'buddy'— the doctor for your Hermes Agent.

Whenever you run into any problem — whether on a VPS, Mac, or PC — you can ask Codex to fix it:

1. A problem happens
2. You ask Codex to fix it
3. Codex fixes it
4. You apply the fix and move on

Why This Matters for Newcomers

This makes it incredibly simple for anyone new to tech. You don't need to debug things yourself — you have a developer buddy that fixes stuff as you go:one brain, two roles.

Key Advantage: Control from Your Phone

• You can control Codex from your phone through theChatGPT app.
• This lets you be anywhere in the world and control your computer with Codex if something breaks.

A New Codex Update

There's a newer Codex update that lets Codex securely use apps on your Macfrom your phone, even when your Mac is locked and the screen is off. This is huge:- For one payment (the hardware), you keep everything local.- If anything goes down, you can still use Codex from your computer to fix Hermes.

Setup Steps

1. Go toCodex by ChatGPT(link in the video description).
2. Download it for Mac, Windows, or your VPS.
3. Log in with your ChatGPT account.
4. You're good to go.

In Practice: The Telegram Setup Snag

During setup, Hermes once stopped and wouldn't restart while configuring Telegram. The fix was simple:

• Open a new Codex chat
• Say, "My Hermes agent is not working, it keeps restarting but not coming back. Can you fix it?"
• Codex fixed it, and Telegram connected successfully.

Expect some difficulties at the start. Once the Buddy System is in place, any snags are easy to resolve.

Next up:Installing the Hermes Agent desktop app.

Installation

1. Search forHermes Agent(Nous Research) or use the link in the description.
2. On desktop,downloadthe app.
3. Follow the setup directions.
4. Open the app — it opens like a regular application on your Mac.
5. HitInstall Hermesand let it do its thing.
6. Launch Hermes (you can launch it from anywhere or by saying "Hermes desktop").

Tour of the Desktop App

The Hermes desktop app is excellent, which is why it's recommended to use it on a personal computer:

• New session— start new sessions with a click
• Skills— all your skills in one place
• Channels— all the channels you want to connect it to
• Everything you develop can be managed here

Step 1: Set Up Your Model Provider

1. OpenSettings.
2. Set up your provider for models — chooseOpenAI Codex.
3. Authenticate with your ChatGPT account (one-click via OAuth).
4. Copy the code, paste it into your ChatGPT account, and hit continue.
5. This sets upGPT 5.5as your primary model.

Step 2: Add a Backup Brain (NousPortal)

Set up asecondary/fallback modelin case you run out of credits or ChatGPT goes down.NousPortalis the company behind Hermes Agent. Using it as a backup gives you:

• Immediate access to all the top-tier models
• No need for 200 different subscriptions (including image and audio models)
• A backup brain if anything is down

To set up NousPortal:

1. Go to Settings → Model.
2. ClickNousPortaland connect it in one click (make sure you have an account).
3. You can switch between NousPortal and OpenAI Codex.
4. Enablefast mode(High is recommended as the best balance).
5. Add a reference model as fallback (e.g., Claude Opus or GLM).

Test It

• Go toNew Session.
• Select your model (bottom right, e.g., GPT 5.5 Fast High).
• Say hello — it responds, and you're ready to work.

Next up:Enabling Computer Use.

What Is Computer Use?

Computer usemeans giving Hermes control of your computer — it becomes 'its own computer.' This makes everything Hermes does significantly simpler, because you can just talk to it and it handles the rest.

How to Enable It

1. In a new session, ask:"Hey, can you help me set up Hermes computer use?"
2. Hermes installs the necessary package so it can control and access the computer when needed.
3. Hermes will ask for permissions — grant them.
4. Specifically, grantKUA driver access(this is optional but enables full potential).
5. HitEnableandQuit & Restart.

Troubleshooting Example

If Hermes says it's not fully done:- Take a screenshot of your settings showing the access you granted.- Send the screenshot to your Hermes agent.- It will verify that permissions are in place.

Once enabled, Hermes feels like it has a computer of its own and can act with a mind of its own.

Note:This is optional but recommended to use Hermes to its fullest potential.

Next up:Tour of a real Telegram setup and connecting channels.

Why Channels Matter (Measure Helpfulness)

Channels let you use Hermes while on the go. The most effective way to measure the value of your AI agents:

1. Amount of work done while you're away from your desk— without channels, Hermes can't get much done while you're away.
2. Percentage of work automated— e.g., Hermes created and deployed a dashboard on its own.
3. Revenue per employee— treat Hermes as an AI employee, measuring your revenue per employee.

A Real Telegram Setup Tour

Samin's actual setup includes:- Asolo channelwhere Hermes and he just talk.- Acapture channel— dump in cool Instagram/YouTube videos, and Hermes ingests and remembers them without responding.-Telegram Topics— topic chats for whatever you're thinking/doing (trading, projects, etc.). For example, a trading topic that sends updates on positions and dashboards.

Example: A trading bot bought a LEAP on Uber and built a dashboard showing a 33% overall return.

The Channels

Telegram

• UsesBotFatherandUserInfoBotto create a bot token.
• Set up/newbot, choose a name/username, copy the token, paste it into Hermes, and save.
• Pair the bot with a pairing code and set ahome chatfor random updates.

Telegram Topics (Multi-Thread Workflows)

• Create a group, add Hermes as an admin.
• EnableTopics(as tabs).
• Ask Hermes to create multiple topic threads (e.g., for a course, trading, and a capture channel).
• Hermes shows a 'typing' indicator so you know when it's working, and you can hop between threads to run things simultaneously.
• Use it from yourphoneto stay mobile.

Slack

• Slack works like anAI teammateintegrated with your team.
• Create a new app from scratch, add scopes (commands, chat:write, chat:read).
• Install the app, get abot tokenand anapp-level token(socket mode).
• Whitelist users so teammates can talk to the bot, or set 'allow all users = true'.
• The bot has context on the whole team, helping avoid siloed information.

iMessage (via Photon)

• UsesPhoton.codes— free to start; create a project.
• Because Hermes has computer use enabled, you can paste the docs and credentials to Hermes and let it configure itself.
• Scan a code to enable texting the number, then message it directly.

Next up:Identity files — personality and memory.

Making Hermes Pleasant & Personal

Now that Hermes is connected to multiple channels, you want it to know you and behave pleasantly over time. This is done throughidentity files.

The 'Soul' — Personality

Hermes' personality comes through itssoul. This is a file (soul.md) loaded each time it talks to you, telling Hermes:- How to speak- What tone to use- Its purpose

The soul evolves over time if you ask it.

Example: Hermes picked up the phrase "airport goblin" and used it naturally later. It talks in lowercase, gives quick links, and avoids the 'ChatGPT droning on for pages' style.

Where the Soul Lives

• Open Hermes → bottom left shows 'default' →Manage Profiles→ this is the soul.
• You don't have to touch it, but you can customize it.

Two Easy Ways to Customize

1. Interview method:Copy the interview questions from the course board and say"interview me."Hermes goes back and forth learning all your details.
2. Paste a soul.md:Use Samin's soul.md (which leads with answers, has opinions, uses evidence) — paste it into the profile's soul.md and save.

The 'User.md' Memory File

Theuser.mdis an identity tag — a file where Hermes writes everything about you. It's not a generic chatbot; it feels like working with an assistant that remembers you across days and weeks.

• The easiest way is to let the agentinterview youand answer all the questions.
• Dump the answers in and say "remember all this stuff."
• Your Hermes then remembers you across days and weeks.

Next up:The challenge of memory and Honcho shared memory.

Why Memory Isn't Simple

Most people think memory is just "the assistant remembers what you said." That's the basic version — and most memory products can do it. Butreal agentic memoryis harder, with four key problems:

1. Single-Session Memory

• A long conversation where the assistant must remember what the user said earlier.
• If you say "keep this direct, no fluff" and then talk for 40 more minutes, it should still follow that.
• It must also remember what the assistant did (created a file, made a plan, promised a follow-up).

2. Knowledge Updates

• The assistant should not talk about your old job — it should know your life has changed.
• Preferences change (last month you liked long explanations; now you want direct scripts).
• Good memory updates and doesn't treat old facts as equally true as new facts.

3. Multi-Session Memory

• Useful info is spread across different conversations (Telegram, Slack, Claude Code).
• A good memory system connects all these pieces, otherwise every new session feels like starting over.

4. Temporal Reasoning (Hardest)

• The assistant must understand time — what happened last year vs. yesterday.
• Example: You liked ice cream last summer, then disliked it, then liked sundaes. It must know which preference matters now.
• Real memory means knowing not justwhatthe user said, butwhenthey said it, whether it's still true, and whether a new fact replaces it.

Why Memory Matters

"Memory is what lets the agent stop acting like a new intern every time you open the chat."

This is the problem solved withHermes + Honcho.

Next up:Introducing Honcho — one shared brain.

What Is Honcho?

Honcho is one of the fastest-growing open-source projects. It solves thecold-start problem: normally a new chat forgets who you are. With Honcho, you become thememory router— you don't have to do anything with memory, and regardless of which chat you're on, it always has context.

One Shared Brain

With Honcho,one shared brainstores everything across all your tools:- Claude Code- Codex- OpenCode- Hermes- OpenClaw

Whatever tool is best for the job, use it — everything feeds into one growing shared brain.

Diachronic Identity

Honcho createspeer cardsfor everyone in your life (and your AI agents).

Example:- You, your manager, and a coworker 'Bob' each have relationships.- Bob is non-technical and can't code.- Your manager thinks of you as hardworking.- Honcho knows all these distinct relationships based on context and saves them.

It ingests conversations from everything, processes them, and the output is much better.

How It Works: 'Dreaming' & Neuromancer

Just like you internalize a busy day while sleeping, Honcho takes raw chats and reasons over them to create useful memory. It does this throughNeuromancer— a fine-tuned model:- Parses all AI messages- Handles knowledge updates- Draws logical implications- Resolves contradictions- Manages peer cards (e.g., 10 different people → 10 peer cards)

Why Hermes + Honcho Beats Standalone Agents

"This is the reason we use Hermes Agent plus Honcho as an AI employee — it works 10 times better than Claude and Codex alone."

A Word of Caution: Honcho Is Powerful (and Dangerous)

Giving all your details to one memory system from everywhere is useful but must be handled responsibly. The good news:- It isopen source.- You canself-hostit on your Mac Mini, so nobody has access to it but you.- You get a great working system that's private.

Next up:Connecting Honcho via OAuth.

Setup Is Simple

Honcho comesprepackaged with your Hermes Agent.

Steps

1. Go toSettings.
2. Go toMemory & Context.
3. UnderMemory Provider, chooseHoncho.
4. ClickConnect via OAuth.
5. Make a Honcho account.
6. Enter your name andAuthorize & Write.
7. That's it — you're connected.

Once connected, your Hermes Agent becomes significantly better at memory.

Next: Management & Observability

With memory in place, the next big question ismanagement— how to manage multiple agents and your team without losing track of goals. That's whereMulticacomes in.

Next up:Multica — chat plus a project board.

Why a Project Board?

Humans like to work in collaborative chat (like Slack). But big organizations work through aproject boardto keep track of everything. On a normal team, everyone works in private DMs and nobody wants to update the board.

If the work was on a board, it's easy to manage.Multica is a free service that lets your AI agents do all the updates for you, so your team and agents work from one place.

Tour of a Real Multica Setup

Samin's Multica includes:-Runtimes— everything he subscribes to (Claude, Codex, Hermes, OpenClaw), managed from one place, with info on where each runs (MacBook, Mac Mini, team computers).-Projects— work that Hermes is leading, with a project manager.-Squads— Hermes is his chief of staff, using Codex and Claude to get work done.-Issues— all the tasks it's worked on, with comments.-Workspaces— separate areas for life vs. business (e.g., 'Hermes course' vs. 'personal').

This givesobservabilityinto your agents — what they're doing — without losing track of millions of threads.

The Key Concept: Hooks

Ahookis an automatic trigger. When Hermes does something (sends a message, starts/finishes work), the hookalways fires deterministically— 100% of the time, and it won't hallucinate.

Two hooks to set up:1.Honcho— every time Hermes finishes a message, it automatically loads your memory.2.Multica— every time you ask it to make something, it updates the board / creates an issue / updates the project.

Result: your board is always managed by Hermes, and you have anagentic project manager from scratch.

Setting Up Multica

1. Go tomultica.ai→ Get Started → log in (Google) → download to your computer.
2. Open it — it auto-detects your tools (Codex, Hermes).
3. In Settings → API Tokens, create a new token (e.g., 'project manager').
4. Give the token to Hermes and ask it to set a hook so every request creates/updates a Multica issue.
5. Hermes creates the project and moves tasks across the board to 'Done'.

Verify: Ask Hermes "If I ask you to create a website, will you create a Multica issue first?" It says yes — because the hook is set.

Even if you use a VPS, Multica shows different computers and the agents on them, and you can control any of them from anywhere.

Next up:Profiles — one Hermes, many AI employees.

Profiles: One Hermes, Many AI Employees

Profilesturn one Hermes desktop setup intodifferent AI employees. Instead of mixing personal, business, and client work, each profile gets its own:- Role- Model- Memory- Skills- Tool access

Examples

• One personal agent
• One for business ops
• One for each client

Each has its own soul, model, connectors, and more. You can clone profiles.

Creating a Profile

1. On the bottom left, clickManage Profiles.
2. HitNew Profile, name it (e.g., 'personal').
3. Clone from default (keeps existing stuff) or chooseNoneto start fresh.
4. Reload — you now have separate profiles, each waking up independently.
5. Set each profile's personality, model, and speed.
6. Create a Multica API token per agent so each shows up in your runtimes.

Squads: Building an Agent Team

Squadslet different models/profiles work together under a leader.

• Go toSquads→ create a new squad (e.g., 'Team Cloud Club').
• Select aleader(e.g., chief of staff Hermes).
• Add members (you, the personal profile, etc.) as other employees.
• The leader can assign tasks to members.

Agent Details & Instructions

Each agent can have instructions forwhen to use what model:-Claude— better at design, planning, product design, high reasoning.-Codex— better at coding / implementing Claude's plans.

Samin didn't write these instructions manually — he just told Hermes: "Set up my team so I'm using different models for different things."

Squad members can also have their ownskills and environment variables. For example, Codex has many YouTube skills, and Hermes can call Codex when it's the best model for the job.

Staying Organized

With Multica, you can create tasks for life and business, track business follow-ups, a YouTube pipeline, and even personal errands (order vacuum parts, home maintenance). You interact with your AI agents as people — you can assign tasks and chat with them directly.

Next up:Connectors — MCPs vs CLIs vs APIs.

The Building Blocks

Hermes isn't magic — it can't access external apps (Gmail, Google Slides) or build automations without you giving it the right access first. That access happens throughconnectors, which can be anMCP, aCLI, or anAPI. These are just different ways to communicate with external applications and use their functionality.

Composio: The MCP Gateway

Composiois the easiest way to get started with MCPs. It says:"Your agent decides what to do, and they handle the rest."

• Your Hermes needs to access Google Calendar, Gmail, and many apps.
• Composio serves as thegatewayto all these applications.
• It makes authentication easy (usually the hard part).
• It hashundreds of apps— connect in one click.

To connect Composio:

1. Sign into Composio (free).
2. Connect your apps (e.g., Google Calendar) with one click.
3. In Hermes, ask it to set up the Composio MCP with OAuth and give you the auth link.
4. ClickAllow Access.
5. Now ask Hermes to read/manage your calendar from any channel, even your phone.

Google Workspace CLI

Google Workspace CLIis a fantastic CLI that lets you use Google Drive, Google Slides, Google Docs, and Google Chat with your Hermes agent. It's easier to use and has a lot of functionality.

To install: take the GitHub link and say,"Hey, can you help me install the Google Workspace CLI?"Hermes walks you through the steps.

MCPs vs CLIs vs APIs

These are all justdifferent ways to communicate with external apps. Learning the difference (with stories and analogies) turns non-technical people into experts.

Next up:Cron jobs — autonomous autopilot use cases.

What Are Cron Jobs?

Cron jobsare how you get your AI agents to do work for youautonomouslyon a schedule.

Example Use Cases

1. Weekly Client Updates

Pull reports from Slack, Drive, Google Docs, and project boards; generate a weekly report; and automatically send it to clients.

2. Niche Research & Content Ideas

Search your niche for new content and trends, and get content ideas every day.

3. Competitor Monitoring

Scrape competitors and monitor prices. On any change, alert you or make changes on your behalf.

4. Invoice / Payment Checker

Connect to Stripe or QuickBooks; if there are unpaid invoices, draft follow-ups and have Hermes send them.

The Key Idea

Once you can connect apps and ask Hermes to do things, you can buildautopilotsthat run on schedules with zero manual work.

The Morning Briefing (First Build)

The first thing to build is amorning briefing— a daily autopilot that sets you up for your day:

The Prompt

"Every morning at 8:00 a.m., create a morning briefing for me. Use my calendar to show meetings. Do research on what I should say. Use my Gmail to find urgent emails from the last 24 hours. Give me an overview of what my team is talking about in Slack. Include today's schedule, urgent emails, important docs, follow-up tasks, and anything to prepare before meetings. Also use Twitter to find trending AI topics."

The 'Biggest Tree Tomorrow' Hack

Every night, ask Hermes:"What's our biggest priority tomorrow?"Getting into this habit produces huge progress toward goals.

The agent research, curates, creates, reviews, and delivers — all autonomously. This is anagentic workflow.

Next up:Packaging workflows into reusable skills.

Skills: Packaging Your Playbooks

Askillis a playbook of one thing you like done a certain way. Even if the research differs each day, the output format stays the same. When you have a process you like and want Hermes to keep doing it,turn it into a skill:

• Easier to reuse
• Easy to share with others and your team
• Hermes comes pre-built with many useful skills
• Having more skills isn't a problem — there's an index, so it's efficient

To make a skill from a cron job: "Hey, can you make this morning briefing into a skill?"

Agentic Workflows

Anagentic workflowis not one prompt — it's acycleof different things happening. It's called anautopilotin Multica.

Hermes does the research, curation, creation, review, and delivery — without you saying anything. This is the most important concept:how can I get my agent to keep doing something useful without me asking?

The 5-Step Recipe

For any workflow:1.Agent— you need an agent like Hermes.2.Connectit to the necessary apps (Google, Chrome, Slack, etc.).3.Massageit until the output is good.4.Set it into a loop(cron job / autopilot).5.Make it into a skill— a package you can share with anyone.

Other Agentic Workflow Ideas

• Executive assistant loop— at night, Hermes asks what coding/project work to run overnight while tokens are cheaper and models more available.
• These are all forms of the same pattern.

Where Everything Is Saved

Everything you do is saved intwo places: Hermes desktop (skills, cron jobs) andMultica(so you can manage from desktop or web).

Next up:Build #2 — an animated website with the /goal command.

The Goal

Build apremium, animated website— with moving sections, animated backgrounds, graphics, photos, and videos — usingone command.

Why Use /goal?

Instead of one-shotting with a normal prompt, use/goal(or/goal).

The goal gives the agent anoutcomeand makes it loop until the job is done:- Not just "make me a website"- But "make a beautiful animated Hermes website that explains the offer, creates visuals, has strong copy"

The agent builds, checks its own results, fixes what's missing, and keeps going. It also needs to make photos, check them, insert them, make videos, insert them, etc.

Tools Needed

Higgsfield AI

• Generates photos, images, websites, and even deploys.
• Install theHiggsfield CLI / MCPand authenticate with OAuth.

Firecrawl

• Scrape your site, monitor it, and visit competitor sites for inspiration.

here.now

• Free, instantweb hostingto deploy apps, websites, videos, and photos.
• Example use: trading bot builds a here.now dashboard each time it responds, so you can see results on your phone.

The /goal Prompt

/goal Build me a website. Make it animated with an animated background and good motion.
The offer: Hermes agents as 24/7 AI employees for content and lead-gen systems.
It can create Instagram carousels, reels, YouTube videos, clips, ads, and scrape competitors with Firecrawl.
Make it feel premium, animated, black-and-white like Notion. Find the Hermes logo and use it everywhere.
The call to action is to book a call. After building, give me the here.now deployed link.

The Result

Hermes built and deployed apremium animated landing pagein one shot:- Animated background- Copy written by Hermes ("Hire Hermes as a content employee")- Generated graphics and images via Higgsfield- Book-a-call CTAs everywhere- All images/videos/audio stored in your Higgsfield account

Key takeaway:Make sure you have Higgsfield (for visuals) and here.now (for deployment) installed so Hermes can make aesthetically pleasing, deployable sites.

Next up:Build #3 — Property Marketing Kit.

The Offer

Real-estate listings often have poor photos. With AI, you can generate afull property marketing kitto make listings feel like somewhere you'd actually live.

What a Marketing Kit Includes

• Floor plans turned into designs— envision the actual layout
• Generated photos of rooms— staged and furnished, so buyers can feel inside
• Showcases— turning barren empty space into an inviting home

Example: A barren empty room becomes a furnished living space, and an empty field becomes an appealing property. This makes listings sell faster.

Pricing

• Sell from$250 to $1,000 per listing(depending on size).
• Hermes can run on autopilot: find new properties, build kits, and send them to realtors.

How to Build It

1. Ask Hermes to find properties in a location (e.g., New York).
2. Install theproperty marketing kit skill(from the course's skills vault).
3. Give it an example property to build a kit for.
4. UseFirecrawlto scrape a list of other properties and their realtors (find email/phone + property link).
5. Make it agoalbecause it takes time.

The Result

• Hermes built the example property's marketing kit (source photos + generated photos).
• It also built kits for additional properties (e.g., 'Clear Water Beachfront') with vignettes and polish.
• It returned a list of other properties with their contacts via Telegram.
• It asked:"Would you like me to generate full marketing kits for all these listings and start sending to these contacts?"

How to Sell It

• Build the kit first and send it for free.Don't ask for anything.
• If they want one for every listing, or the whole brokerage, scale it up.
• Put it on autopilot with Multica — find new realtors, make kits, and keep sending.
• You're sellingtwice the interest on your listing in 7 daysdue to much better photos and videos.

Next up:Build #4 — Instagram thread carousels.

The Common Content Core

Across platforms, there's a common content format —carousels(reels/carousels on Instagram and TikTok, carousels on LinkedIn). All platforms follow the same workflow:

1. Researchtrends, keywords, and what your ICP would be educated by
2. Writethe script in a specific format
3. Create / film
4. Edit
5. Post
6. Analyze

Types of Carousels

1. Simple / 'Twitter-style'

• Plain but high-converting
• Cheap — no image model or lots of tokens needed (mostly code + research)
• Great for when something new comes out

2. Designed / Pixel-Style

• Nice graphics and design
• UsesHiggsfieldto generate images
• To match any style, take a screenshot of inspiration, insert it into Hermes, and ask it to make a skill in that style

How to Create Carousels

Using theInstagram thread carousel skill:

"Hey, [model update] just came out. Can you create a carousel about it using this Instagram thread carousel skill?" (Paste your headshot.)

The skill handles most of the heavy lifting. Generated content is stored inartifacts.

For designed carousels, use thepixel-style skilland Higgsfield for images.

Posting on Autopilot

Two ways to post automatically:

1. Composio— connect Instagram, TikTok, Facebook, YouTube, etc., and have Hermes post directly.
2. Blotato— a content calendar tool that posts across all socials and gives a calendar-style view. Copy the MCP config into Hermes.

Example: "Schedule one carousel every day." Hermes creates a content calendar and posts across all your socials.

Putting It on Autopilot

• Set a cron job or Multica autopilot to generate content daily.
• Hermes does the research and writing; you just approve.

Next up:The content repurpose octopus & AI video editing.

The Content Octopus Concept

The highest-leverage way to grow a brand is to makeone long-form piece(like a YouTube video) and turn it into everything else:- Clips- Reels- Articles- Tweets- Community posts- Carousels

How It Works

Input:a YouTube video + a skillOutput:all the repurposed content

Using thecontent repurpose skill:

"Download and install this skill, then use it to make a dashboard of the content repurpose system and activate it on this video."

Results

• Clips— short-form clips
• Tweets(8 tweets)
• LinkedIn postsand lead magnets
• A long-form blogfor SEO
• X threads
• Short-form reel scripts(with new hooks for refilming)
• All copy isfree of AI slopbecause it's run through a loop

AI Video Editing (Two-Pass)

Editing with AI happens intwo phases:

Pass 1: Cut

• Put the raw clip through
• Cut out bad takes, ums, ahs, and pauses

Pass 2: Add B-roll

• Find and insert B-roll into one final polished video
• Use two skills: acut-video skilland aB-roll / motion-graphics skill

Example Workflow

"Edit this reel. I have a lot of bad cuts. Make a goal that you will give me the final output video. First pass: use the cut skill to cut out ums and bad takes. Second pass: use the B-roll skill to find and insert B-roll, with me at the bottom and B-roll on top. Attach clips of me from my YouTube video and edit those in."

The result: a polished reel with captions, bad takes removed, and B-roll inserted.

The Payoff

FromoneYouTube video, Hermes made: clips, edited clips, Instagram reels, carousels, articles, tweets, and posts. This is one of the most powerful and sellable builds in the course.

Next up:Build #5 — E-com ad factory & community manager.

E-com Ad Factory

If you have a product (not talking-head content), Hermes can makeimage and video adsfor it — includingputting you in the adsusing your Higgsfield soul.

How It Works

1. Take a picture of your product.
2. Install thee-com ad skill(which researches competitor ads).
3. Give Hermes the product photo and your Higgsfield soul.
4. Hermes creates image and video ads with benefit-driven copy, styled visuals, and your likeness.

Results

• Beautiful, benefit-driven image ads
• Video ads with the creator in them
• Copy like "Emotional support" and contrast-based creative
• Set a cron job to make new ads on autopilot
• Connect theMeta Ads MCPto run and test which ads perform best
• Essentiallyhundreds of ads a day, all managed by Hermes

Community Manager + AI Inbox

Once you have clients and a community, you need customer support.

Community Manager (Skool Bot)

• Hermes lives inside your Skool/Discord community
• Answers questions, moderates, and gives sentiment reports
• Builds cool dashboards and prompts for people

How to Build It

1. Make a dedicated community account for the bot
2. Give it credentials and set what it should do
3. Setcron jobs(e.g., 9 a.m., 2 p.m., 5 p.m.) to check posts and comments
4. It runs as a24/7 co-pilot: morning check, midday reply sweep, evening community summary
5. Weekly it learns from the community and gives you a report

AgentMail: Giving Your Agent Its Own Email

AgentMail.togives your agent an inbox:1. Log in and create an inbox for your agent (e.g., test-hermes123@domain)2. Create an API key3. Give the credentials to Hermes and ask it to onboard and send a test email4. Hermes can now review, summarize, draft, organize, follow up, and turn things into tasks

Example email from Hermes: "I have my own email address now. If you ever need anything done, you can email me here or forward me something."

Next up:Self-improving trading bot.

Why Hermes Makes a Great Trading Bot

Hermes'self-improvement loopis perfect for trading:- Teach tasks and your algorithm over time- Detect and learn patterns (buy/sell points)- Reads your past trades and improves- The compound effect makes your bot better over time

Three Steps to Build a Trading Bot

Step 1: Give Access to a Trading Platform (Co-Invest)

• Go toco-invest.ai(free account)
• It supports stocks, crypto, forex, and even event prediction
• Copy the OAuth and connect the MCP in Hermes/Telegram
• Enablepaper tradingto test safely
• Ask for portfolio status and an HTML dashboard (hosted on here.now)

Step 2: Add a Strategy (QuiverQuant)

• QuiverQuanttracks congressional and famous-person trades (Nancy Pelosi, Elizabeth Warren, Michael Burry, Bill Gates)
• Thestrategies foldershows how well strategies have performed (e.g., 57.8% return over 3 months)
• Copy the QuiverQuantAPI tokenand connect the MCP to Hermes
• Hermes can copy-trade strategies automatically

Step 3: Set It on Autopilot

"Set up a paper trading account and copy the trades of [person]. Start with $10,000. Make a here.now dashboard and give me an update every week. Set a cron job every day during trading hours to check QuiverQuant. Every Friday, give me an update. Make the autopilot in Multica too."

The bot:- Mirrors trades with a disclosure lag- Creates a here.now dashboard showing starting position- Runs daily polls and weekly updates

Summary

You need two services:1. Abroker(Co-Invest) to execute trades2. Astrategy(QuiverQuant) for the playbook

Hook Hermes up to both, and it runs on cron jobs as your trading bot — switch to real money whenever you're ready.

Next up:Health & longevity coach.

The Goal

Knowing what to do is rarely the problem —doing it every day consistentlyis. Hermes becomes youraccountability partnerfor health.

The Three Easy Moves

1. Feed it your blood work— so it knows your biology
2. Talk through a protocol— it recommends a supplement stack and routine
3. Let it schedule— it makes its own cron jobs to remind you daily

The System

• Give Hermes yourblood test(PDF) → it lists supplements and a plan
• Voice memosand yoursleep tracker(e.g., Eight Sleep) feed Hermes your sleep score and reports
• Hermes checks in on workouts, breaks, and supplements
• It nudges you if you're missing something

Example Cron Jobs

• "Did you take your stack?" (supplements)
• "Have you worked out yet?"
• "Maybe step away" if it detects heavy screen time
• Nightly: "What are your next three priorities for tomorrow?"

Results

Samin went from waking up sneezing and struggling to get out of bed, to waking up energized with clear daily direction. His stack:vitamin D, fish oil, vitamin C, creatine.

Key Takeaway

"The best health tracker is the one that tracks you back."

It's private (you own your data) and runs on autopilot. Use voice memos and capture logs — it logs, nudges, and keeps you acting.

Next up:Jarvis — one command center for all agents.

What Is Jarvis?

Jarvis (andAgent Club) is acommand centerthat unifies all your chats and agents in one destination. Instead of chatting with Hermes in one place, Codex in another, and Claude Code in a third, you manage everything from one dashboard.

Jarvis Mode

Jarvis Mode is the most requested feature. It:- Activates on akeyword- Controls your computer (opens apps and tabs)- Speaks back in acloned voice- Plans and starts your day- Reads your calendar and has context on your life

Example Demo

"Good morning, Jarvis. What's my day looking like?""Good morning, sir. I've organized your calendar. At 9:30 we start deep work on the Hermes course. At 11:00 I'm pulling up your Skool community. At 1:15 you have new member intros. You've got a friend's birthday at 7 tonight, so let's wrap around 5. Your main goal today is finishing the Hermes course — three steps left. I've already started the scripts for your Fable video."

Jarvis shows your agenda, top three priorities, and colors/organizes your calendar.

The Five-Part Recipe for Building Jarvis

1. Life context (Honcho)— it knows what's important to you and surfaces your top priorities from your nightly check-ins.
2. Calendar organization— via Composio + Gcal MCP.
3. Computer use— opens tabs and applications.
4. Voice— via Eleven Labs (or Voicebox for local) + Hermes voice.
5. The HUD interface— take screenshots of a vibe you like, define what's important, and ask Claude to design the interface; then use Fable to build it. Use Hermes as anorchestratoracross your squad (Claude for design, Codex for coding).

How to Get It

Building Jarvis takes time. Samin packaged it in his Skool community (Agent Club) so you can grab it prepackaged, and his team helps you set it up if you're new.

Next up:Selling & pricing your builds.

Mark Cuban on the Opportunity

"I've been through every technology evolution, and this blows them all away. When I was 24, I was walking into companies who had never seen a PC and explaining its value. There are going to be integrators... Companies don't understand how to implement AI right now."

• 33 million companies in the U.S.; ~30 million are solopreneurs without AI budgets or experts.
• The opportunity: learn toimplementAI in companies for competitive advantage.

Three Questions Before Selling Anything

1. What can I sell?— Everything you built in this course is a product.
2. What do I charge?
3. How do I approach people?

The Offer Menu

Level 1: AI Employee Install — $1,500 setup

Set up Hermes on their VPS/Mac Mini, connect channels, add memory, track projects, connect Google. Frame it as a24/7 AI employee that learns your businessand is live in 7 days.

Executive Assistant (autopilot) — from $500

Add Composio/MCPs and cron jobs. Morning briefings on their phone. This is thecheapest 'yes'in your lineup — great for founders and executives.

Animated Websites — $250–$500 each

Lead magnet using the /goal command, deployed fast. "Premium animated website that sells your offer, live in 24–48 hours."

Property Marketing Kits — $250 per listing

Scrape realtors with Firecrawl, stage with Higgsfield, email each agent. Build first, send free, then scale to brokerage.

Content Engine (carousels) — easy recurring

Auto-researches trends, writes in their voice, designs carousels, schedules/posts via Blotato. Great for zero-social-presence clients.

Content Repurpose System — $1,000–$3,000/month

For YouTubers/business channels. "Film once a week, it shows up everywhere daily." One video becomes 30.

Reel Editing — pennies/minutes per reel

Cut ums/ahhs, add B-roll, deliver captioned reels in 24 hours. "Your first one's free."

E-com Ad Factory

Computer ad research + image/video ads, managed via Meta Ads MCP. "Fresh new creatives" is the offer.

Community Manager — from $500 (or per resolved incident)

Bot in Skool/Discord with email, checking on cron. "Every member answered within 24 hours in your voice."

Premium Builds

• Trading bot(Co-Invest + QuiverQuant)
• Longevity coach— accountability for vitamins/gym
• Jarvis / Agent Club— performative demo that closes the call

Pricing Principles

1. Price the outcome, not hours.
2. Always build first— make it work for yourself before selling.
3. Anchor to the human alternative— what it would cost to pay people, and how much more scale/volume you get.
4. Setup + retainer is the way to go— setup frees your time; retainer keeps you on and stacks over time.
5. Always have a risk reverser— take the risk off the client, especially for your first sale. E.g., "If you don't love it, you don't pay."
6. Send free stuff first— demo your work, then offer more.

What to Do Next

• Build these systems for yourself first.
• Use them, refine them, and prove they work.
• Then reach out, send free value, and start selling.
• If you get stuck, the community (Cloud Club / Agent Club) is there to help.

You now have the tools to build an agent-run company and sell these systems for recurring revenue.