Install Hermes Agent in 15 Minutes: Complete Setup Guide

Hermes Agent installs in one command. Seriously. You’ll spend more time opening your terminal than waiting for the installation to finish.

No Docker complexity. No multi-step wizards. No “run this, then this, then debug this” nightmare. One line, and it’s running.

Let’s do this.

Before You Start: Check Your Specs

You need:

4GB RAM minimum (8GB recommended, 16GB+ if you’re running local models)
2GB free disk space (more if using local LLMs)
Linux, macOS, WSL2, or Android (Termux)
Bash shell (standard on all of these)

Don’t have these? You can still run Hermes on less powerful hardware—it’ll just be slower. The 4GB is a practical minimum for local LLM inference. If you’re only using cloud APIs, 2GB RAM is fine.

Step 1: Run the Installation Script (One Command)

Open your terminal and run:

curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash

That’s it. The script will:

Clone the Hermes repository
Install dependencies (Python, required libraries)
Create the .hermes directory in your home folder
Set up your PATH so the hermes command works globally

On most machines, this takes 2-5 minutes depending on your internet speed.

What it installs:

The Hermes Agent CLI tool
Python runtime (if not already installed)
Configuration directories

After it finishes: You’ll see:

✓ Hermes Agent installed successfully
✓ Run 'hermes --version' to verify

Step 2: Verify Installation Worked

hermes --version

You should see something like:

Hermes Agent v0.8.1
Built with ❤️ by Nous Research

If you get “command not found”, you might need to restart your terminal or add the installation path to your shell config. (This is rare, but we’ll cover it in the troubleshooting section.)

Step 3: Initialize Hermes

Now run the setup wizard:

hermes setup

This interactive script will ask you:

1. Choose your LLM provider

Options:

local — Use Ollama (free, private, see Article 9)
openai — Use OpenAI’s API ($)
anthropic — Use Claude’s API ($)
openrouter — Unified API provider ($, but supports many models)

Pick based on your preference:

First time? Choose local (requires Ollama installed, but free and fast)
Have an OpenAI key? Choose openai
Want to test with local? You can change this later

2. Configure your provider

Depending on what you chose:

If Local Ollama:

Provider: local
Ollama endpoint: http://localhost:11434
Model name: mistral

(We’ll install Ollama in Article 9. For now, skip if you don’t have it.)

If OpenAI:

API Key: sk-xxxxxxxxxxxxxxx  (paste your key)
Model: gpt-4

If Anthropic:

API Key: sk-ant-xxxxx
Model: claude-opus-4-7

3. Choose your first platform

Options:

cli — Just use the terminal (easiest for testing)
discord — Connect a Discord bot
slack — Connect a Slack bot
telegram — Connect a Telegram bot
skip — Set up later

For your first time: Choose cli. It’s instant and lets you test without any bot tokens.

The setup wizard will create a config file at ~/.hermes/config.yml. You can edit this manually later, but the wizard handles 99% of what you need.

Step 4: Start Using Hermes

After setup, just run:

hermes

You’ll enter interactive mode. You can type questions directly:

You: What is Hermes Agent?

Hermes: Hermes Agent is a self-improving autonomous AI assistant...

Type exit or quit to leave.

That’s it. You’re running Hermes.

What Just Happened

You now have:

✅ Hermes Agent CLI installed globally
✅ Configuration stored in ~/.hermes/
✅ An LLM provider configured (local, OpenAI, or Anthropic)
✅ Ability to chat directly via CLI

The memory system is already working. Everything Hermes learns is stored in ~/.hermes/memory/.

Next Steps: Connect a Platform (Optional for Now)

You can chat in the terminal forever if you want. But Hermes gets more useful when it lives on Discord, Slack, or Telegram.

We’ll cover that in Article 4. For now, you have a working Hermes instance. Test it out.

Customizing Your Setup (Intermediate)

If you want to tweak settings without running the wizard again, edit your config:

nano ~/.hermes/config.yml

Common tweaks:

# Increase context window for longer conversations
llm:
  context_window: 8192  # Default is 4096

# Enable streaming responses (faster feeling)
llm:
  streaming: true

# Set inference parameters
llm:
  temperature: 0.7     # Creativity (0=deterministic, 1=creative)
  top_p: 0.9          # Diversity of responses

Save and restart Hermes for changes to take effect.

Hardware Considerations

Do I need a GPU? Not required. Hermes works fine on CPU (slower, though).

Can I run on a laptop? Yes. It’ll use some memory, but totally viable.

What about a Raspberry Pi? Technically possible, but very slow. Better to use cloud LLM provider.

Should I use a VPS/cloud server? Yes, if you want it running 24/7 without your laptop. Install the same way, configure via SSH.

Troubleshooting Installation

“curl: command not found” You’re on a system without curl. Install it:

macOS: brew install curl
Ubuntu: sudo apt install curl
Fedora: sudo dnf install curl

“Permission denied” on install The script doesn’t need sudo. If you’re getting permission issues, you might have a PATH problem. Try:

export PATH="$HOME/.hermes/bin:$PATH"
source ~/.bashrc

“hermes: command not found” after install Restart your terminal, or run:

source ~/.bashrc  # Linux/WSL
source ~/.zshrc   # macOS (if using zsh)

Installation hangs at “Installing dependencies” Your internet might be slow. Wait 5 minutes. If it still hangs, cancel (Ctrl+C) and try again.

“Python not found” error Hermes needs Python 3.8+. Install it:

macOS: brew install python3
Ubuntu: sudo apt install python3
Windows: Download from python.org

Verify Everything Works

Run this command to check your installation:

hermes status

You should see something like:

Hermes Agent Status:
✓ Installation: OK
✓ Configuration: OK
✓ LLM Provider: Connected (openai)
✓ Memory: Ready
✓ Platforms: 0 connected

If everything says ✓, you’re good to go.