Install Hermes Agent: desktop app or terminal

You can install Hermes Agent in two ways: with the desktop app for Windows and Mac, or with one command in the terminal. You do not have to be a developer. In this guide, I show you the desktop app first and then the terminal version, with commands you can copy and paste. Both routes lead to the same agent, with the same memory and capabilities.

The examples come from our own Hermes installation at Voltti. The installation output and startup screen were assembled for this explanation; what you see depends on your version and configuration.

Installing Hermes through the desktop app or terminal, with a penguin as your guide
Two routes to the same agent: the desktop app or the terminal.

Why two routes? The desktop app is the quickest start if you rarely open a terminal: download it, open it, and type your first question. The terminal version gives you more control. It is also the starting point if you later want to run the agent on a separate computer or server. Whichever you choose, the Hermes documentation explains each step in detail.

What do you need before you start?

For the desktop app, you need Windows or a Mac with Apple silicon. Hermes does not support Macs with an Intel processor, not even through the terminal. The official installation page says it plainly: “The macOS installer is Apple Silicon only.”

The terminal version runs on Linux, macOS, and Windows, through WSL2 or PowerShell. On Linux and macOS, you need Git and curl, among other tools. The installation script sets up Python, Node.js, and the other components itself. You also need an AI model, through an account with a provider or your own API key. In the desktop app, you can start with a free trial model first.

Install Hermes Desktop without a terminal

The desktop package includes everything the agent needs, including Python and the other components. You do not need to install anything through a screen full of terminal commands. Here is how it works:

  1. Download the package for your system from the Hermes website.

  2. Open the file. On Windows, use App Installer. On a Mac, open the disk image and drag Hermes to the Applications folder.

  3. Start Hermes. A short guided chat helps you with the first settings and runs on a free trial model.

  4. Then choose the model you want to keep using.

  5. Type your first question, and the agent responds.

For that model, you can sign in to a provider such as Nous Portal. One account then covers the model and built-in tools such as web search and image generation. If you already have API keys from OpenRouter or Anthropic, for example, you can enter those instead.

From then on, you use the app for conversations, files, and memory. We explain how to work together as a team in our article about Hermes for teams.

When should you choose the terminal version?

Choose the terminal version if the agent should stay on, even when your laptop is closed. The desktop app runs on the computer you are using. Through the terminal, you can also run the agent on a separate small computer or server. Our article on when you need a Linux computer for an AI agent explains when such a separate machine makes sense.

You can then connect the agent to Google data, as explained in our article about retrieving marketing data with an AI agent. The agent can also configure that connection from Hermes Desktop if the necessary tools and access are available.

You can copy the commands below and paste them into your terminal. Wait for each step to finish before continuing. If something fails, the installation script tells you what is missing. Often, one extra command suggested on screen resolves it.

installing hermes on Linux

$ curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash

Hermes Agent Installer

→ Detected: Linux (arch) ✓ Managed uv found (uv 0.9.27) ✓ Python found: 3.11.15 ✓ Git 2.51.0 found ✓ C++ compiler found ✓ Node.js v26.14.0 found (Hermes-managed) ✓ Internet connectivity looks good → Cloning hermes-agent repository... → Creating virtual environment... → Installing dependencies...

┌─────────────────────────────────────────────────────────┐ │ ✓ Installation Complete! │ └─────────────────────────────────────────────────────────┘

📁 Your files:

Config: ~/.hermes/config.yaml API Keys: ~/.hermes/.env Data: ~/.hermes/cron/, sessions/, logs/ Code: ~/.hermes/hermes-agent/

🚀 Commands:

hermes Start chatting hermes setup Configure API keys & settings hermes config View/edit configuration hermes update Update to latest version

⚡ Reload your shell to use 'hermes' command:

source ~/.bashrc

Illustrative installation output assembled from the installation script and checks of our own installation. The original terminal text is preserved unchanged. The version numbers come from an earlier installation. The installer now sets up its own Python version, so you do not need a specific one.
installing hermes on Linux

After installation: choose a model, check, and update

The installation script usually starts the configuration wizard itself. To run it again later, use hermes setup. The wizard asks which model you want to use and which tools to enable. Then run hermes doctor to check that everything works. This built-in check tells you whether your configuration is complete.

source ~/.bashrc   # on a Mac: source ~/.zshrc
hermes setup
hermes doctor

On Windows, you can also use the terminal instead of the desktop app. Open PowerShell and run this command from the installation guide:

iex (irm https://hermes-agent.nousresearch.com/install.ps1)

Already installed the command-line version? Run hermes desktop to add the desktop app.

The startup screen below comes from our own installation on version v0.21.1. Hermes gets frequent updates, so your version is probably newer. The process stays the same. Keep a terminal installation current with hermes update; the desktop app updates itself. To move to another computer, first create a backup with hermes backup and restore it on the new machine with hermes import.

At first, give the agent only the access its first task needs. In chat channels, the agent by default responds only to people you allow, and risky commands ask for your approval first. Read more about working safely with an AI agent on our page about security and privacy.

hermes — first launch after installation

$ hermes


██╗  ██╗███████╗██████╗ ███╗   ███╗███████╗███████╗       █████╗  ██████╗ ███████╗███╗   ██╗████████╗
██║  ██║██╔════╝██╔══██╗████╗ ████║██╔════╝██╔════╝      ██╔══██╗██╔════╝ ██╔════╝████╗  ██║╚══██╔══╝
███████║█████╗  ██████╔╝██╔████╔██║█████╗  ███████╗█████╗███████║██║  ███╗█████╗  ██╔██╗ ██║   ██║
██╔══██║██╔══╝  ██╔══██╗██║╚██╔╝██║██╔══╝  ╚════██║╚════╝██╔══██║██║   ██║██╔══╝  ██║╚██╗██║   ██║
██║  ██║███████╗██║  ██║██║ ╚═╝ ██║███████╗███████║      ██║  ██║╚██████╔╝███████╗██║ ╚████║   ██║
╚═╝  ╚═╝╚══════╝╚═╝  ╚═╝╚═╝     ╚═╝╚══════╝╚══════╝      ╚═╝  ╚═╝ ╚═════╝ ╚══════╝╚═╝  ╚═══╝   ╚═╝

┌─ Hermes Agent v0.21.1 (2026.9.7) · upstream ead7e91d ───────────────────────────────────────────────────┐
│                                     Available Tools                                                     │
│  ⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢀⣀⡀⠀⣀⣀⠀⢀⣀⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀     browser-use: browser_exec                                           │
│  ⠀⠀⠀⠀⠀⠀⢀⣠⣴⣾⣿⣿⣇⠸⣿⣿⠇⣸⣿⣿⣷⣦⣄⡀⠀⠀⠀⠀⠀⠀     clarify: clarify                                                    │
│  ⠀⢀⣠⣴⣶⠿⠋⣩⡿⣿⡿⠻⣿⡇⢠⡄⢸⣿⠟⢿⣿⢿⣍⠙⠿⣶⣦⣄⡀⠀     code_execution: execute_code                                        │
│  ⠀⠀⠉⠉⠁⠶⠟⠋⠀⠉⠀⢀⣈⣁⡈⢁⣈⣁⡀⠀⠉⠀⠙⠻⠶⠈⠉⠉⠀⠀     cronjob: cronjob_manage                                             │
│  ⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⣴⣿⡿⠛⢁⡈⠛⢿⣿⣦⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀     delegation: delegate_task                                           │
│  ⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠿⣿⣦⣤⣈⠁⢠⣴⣿⠿⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀     file: patch, read_file, search_files, write_file                    │
│  ⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠉⠻⢿⣿⣦⡉⠁⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀     memory: memory                                                      │
│  ⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠘⢷⣦⣈⠛⠃⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀     process: process_manage                                             │
│  ⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢠⣴⠦⠈⠙⠿⣦⡄⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀     (and 7 more toolsets...)                                            │
│  ⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠸⣿⣤⡈⠁⢤⣿⠇⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀                                                                         │
│  ⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠉⠛⠷⠄⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀     Available Skills                                                    │
│  ⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢀⣀⠑⢶⣄⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀     apple: apple-notes, apple-reminders, findmy +1 more                 │
│  ⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⣿⠁⢰⡆⠈⡿⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀     autonomous-ai-agents: claude-code, codex, computer-use +2 more      │
│  ⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠳⠈⣡⠞⠁⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀     creative: architecture-diagram, ascii-video +8 more                 │
│  ⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀     devops: sdlc-review                                                 │
│                                     email: email-inbox-triage, himalaya                                 │
│  glm-5.3 · Nous Research            media: gif-search, songsee, youtube-content                         │
│  /home/gebruiker                    note-taking: obsidian                                               │
│  Session: 20260909_203509_397644    productivity: airtable, box +12 more                                │
│                                     research: arxiv, competitor-news-monitor +2 more                    │
│                                     social-media: xurl                                                  │
│                                     software-development: codebase-inspection, dogfood, github +9 more  │
│                                     web: blocked-page-recovery                                          │
│                                                                                                         │
│                                     21 tools · 58 skills · /help for commands                           │
└─────────────────────────────────────────────────────────────────────────────────────────────────────────┘

Welcome to Hermes Agent! Type your message or /help for commands.
* Tip: Smart model routing can auto-route simple queries to a cheaper model.

# glm-5.3 | ctx -- [██████████] -- | 1s | 0s
❯
An illustrative fresh-install startup screen: the banner, model and session, tools and skills enabled by default, and the prompt. The original output, including the Dutch example directory name, is unchanged. You can type your questions in English or Dutch.
hermes — first launch after installation

Installation not working?

Start with hermes doctor. This check tells you exactly what is missing and how to fix it.

Does your terminal say the hermes command does not exist after installation? Reload your shell with source ~/.bashrc (on a Mac: source ~/.zshrc) or open a new terminal window.

Do you see a message that no API key is set? Choose your provider and model with hermes model.

Is a setting missing after an update? Run hermes config check and then hermes config migrate.

When a step fails, the installation script shows the last lines and where the full log file is stored.

What can you do with it?

A working installation is the start. From there, the agent develops with your work. It saves agreements in memory and builds skills for recurring tasks. Want to reach the agent through a chat channel your team already uses, such as Slack, Discord, or Telegram? Run hermes gateway setup. Through MCP servers, connections to other tools, the agent can also work with Google Search Console, for example. We explain what that looks like for a business on our page about the Hermes AI agent for businesses.

Not sure which task to give the agent first? Hermes Agent explained covers six business use cases, and our comparison of Hermes with ChatGPT and Copilot shows where they differ.

Prefer not to install it yourself? Voltti installs Hermes and provides support for businesses in Belgium and the Netherlands. See how a pilot works or discuss your needs using the button below.

Portrait of Seppe Gadeyne

Seppe Gadeyne is a developer and the owner of Voltti. He sets up AI agents for companies in Belgium and the Netherlands.

Updated on

Frequently asked questions

01Do I need to know how to code to install Hermes?

No. You install the desktop app without a terminal: download it, open it, and start it. On first launch, a short guided chat helps you get started on a free trial model. For the terminal version, you copy and paste the commands from this article. If something is missing, the installation script suggests the command that fixes it.

02Which systems does Hermes run on?

The desktop app runs on Windows and on Macs with Apple silicon. Macs with an Intel processor are not supported, not even through the terminal. Through the terminal, you install Hermes on Linux, macOS, and Windows, with WSL2 or PowerShell. On Android, you use the Hermes package for Termux. The installation script sets up Python, Node.js, and the other components itself.

03What does Hermes cost?

The software itself is open source and free. You pay for the AI model that powers the agent, through a subscription such as Nous Portal or with API keys from a provider you choose. You select that during configuration. The desktop app lets you start with a free trial model. With a local model, you do not pay per question, but you do need suitable hardware.

04Can I move from desktop to a server later?

Yes. The agent saves configuration, memory, and skills in a fixed directory. Create a full backup with hermes backup and restore it on the new machine with hermes import. That way, you move the installation without starting over. Note that an export of a single profile leaves out keys by design, so it is not a full backup.

05Can I use Hermes without an internet connection?

Installation and externally hosted models need an internet connection. To work fully offline, connect a local model through your own endpoint. The model then runs on your own hardware. Read how that works and when it makes sense in our article about local and hybrid AI solutions for businesses.

Ready to get started with Hermes?

Discuss your pilot