Free starter · 30 min · No signup needed to read

Your first agent.
In 30 minutes.

A focused walkthrough that takes you from a blank terminal to a working AI agent that reads files, summarizes them, and saves the result. No prior agentic AI experience needed. Just a laptop and 30 focused minutes.

Start now ↓ See the full course

00 What you'll have at the end

By the end of this guide you'll have:

  • A working AI agent running on your machine
  • The ability to ask it to read any file and produce a summary
  • A starting point for everything else in the full course
This is real. Not a toy demo. The same architectural pattern scales from this 30-minute starter to the always-on multi-agent Hermes system that runs my life. You're starting at the foundation.

01 Mental model: what is an agent?

Strip away the marketing. An "agentic AI" is four pieces wired together:

Brain

An LLM — Claude, GPT, Llama, whatever. Reads context, decides what to do next.

Tools

Functions the LLM can call: read a file, write a file, run a command, hit an API.

Memory

Where it stores what it's learned — usually files, sometimes a database.

Loop

The thing that keeps it running. Could be a single prompt, a cron job, or a 24/7 process.

That's it. Everything else — multi-agent orchestration, MCP servers, vault patterns, brain cycles — is built on top of these four pieces.

For this starter, your agent will use:

  • Brain: Claude (via Claude Code, the official CLI)
  • Tools: Read, Write, Bash — built in to Claude Code
  • Memory: a file on your disk
  • Loop: a single command you run from the terminal

02 Setup (5 min)

You need three things. If you already have them, skip ahead.

1. A terminal

On Mac: Terminal (built in). On Windows: Windows Terminal with PowerShell or the Git Bash shell. On Linux: you already know.

2. Claude Code (the CLI)

Claude Code is the official Claude command-line tool. It IS the agent — you're not building one from scratch, you're configuring one.

install
# macOS / Linux
$ curl -fsSL https://claude.ai/install.sh | sh
# Windows (PowerShell)
$ irm https://claude.ai/install.ps1 | iex

Then verify:

verify
$ claude --version
claude-code 1.x.x

3. An Anthropic API key (or sign in)

Claude Code needs to talk to Claude. Either:

  • Easiest: run claude for the first time and follow the browser sign-in flow (Claude account)
  • Pro: get an API key at console.anthropic.com and set ANTHROPIC_API_KEY in your environment
Cost reality: Claude Code is metered by tokens. Running a starter agent like this one costs pennies. The full Hermes system, even running 24/7, can hold under $20/mo if you route smartly. We cover cost control deeply in Module 3 of the full course.

03 Build your first agent

Now the fun part. Make a folder, drop a sample file in it, then ask Claude Code to summarize it.

first-agent
$ mkdir my-first-agent && cd my-first-agent
$ echo "The mitochondrion is the powerhouse of the cell. Mitochondria are responsible for energy production through ATP synthesis. They contain their own DNA, separate from the nucleus." > sample.txt
$ claude

Claude Code launches an interactive session. You're now talking to your agent.

Type this prompt and press Enter:

Prompt to your agent
Read sample.txt, write a one-sentence summary of it, and save the summary to summary.md. Then tell me what you did.

Watch what happens. Claude Code will:

  1. Use the Read tool to open sample.txt
  2. Compose a one-sentence summary in its head
  3. Use the Write tool to create summary.md
  4. Report back to you with what it did

Each of those tool uses is the agent doing things in the real world. That's the fundamental shift from "chatbot" to "agent": tools.

04 Run it and see it work

Quit Claude Code (Ctrl+C twice or type /exit). Then:

verify output
$ ls
sample.txt summary.md
$ cat summary.md
Mitochondria are organelles that produce
energy through ATP synthesis and contain
their own DNA.

The summary file exists. Your agent created it. You didn't write that file — the agent did.

That's the magic moment. Sit with it.

05 What you just did

You wired a brain (Claude), tools (Read + Write), memory (the filesystem), and a loop (one prompt) into something that did work.

This is the entire pattern. Every Hermes capability — autonomous research, scheduled briefs, multi-agent coordination, custom skills — is built by extending these four pieces. Not by switching frameworks. Not by buying SaaS. By composing.

The leap most people don't make: they assume agentic AI requires complex frameworks (LangChain, AutoGPT, custom Python). It doesn't. The thing you just did, scaled up, becomes Hermes. The full course teaches you the scaling.

06 Where to go from here

You have a foundation. Three honest options:

Keep going on your own

Try this prompt next: "Read all .txt files in this directory and write a combined summary to digest.md." Watch how the same agent extends naturally to multi-file work.

Join the Discord

Free. The community channel is open to everyone — even if you never buy the course. Post your summary.md in #first-agent.

Join Discord →

Take the full course

The next 9 modules walk you through making this agent persistent, autonomous, customized to YOUR life, and always-on. First cohort starts May 4, 2026. Founder pricing.

See pricing →

That's the starter. If this got you somewhere, tell a friend.

— Alex