Get started with Hermes Agent

This page is about the first real Hermes run. If you are searching for how to use Hermes Agent for the first time, this is the right page. The goal is not to explain every feature. The goal is to get you from install to a useful first conversation with the right mental model.

What you need before you start

Before the first run, make sure you have:

• a working local install of Hermes Agent
• a supported environment such as Linux, macOS, or WSL2
• a provider or endpoint you can use with hermes model
• a small real task in mind

If Hermes is not installed yet, go to Install Hermes Agent.

Step 1: install Hermes

The official quickstart begins with the one-line installer:

curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash
source ~/.bashrc   # or source ~/.zshrc

That gets the hermes command onto your machine and prepares the runtime.

Step 2: configure your provider

After installation, the official docs point new users to these commands:

hermes model
hermes tools
hermes setup

The practical meanings are:

• hermes model chooses the provider and model Hermes should use
• hermes tools controls which tools Hermes can call
• hermes setup walks through the broader setup flow in one place

You do not need to perfect your configuration on the first pass. You only need a working provider and a sane tool setup.

Step 3: start the CLI

The official first-run command is:

hermes

Hermes opens in a terminal interface. Officially, Hermes is a TUI, not a web UI. That matters because the CLI is still the clearest way to understand how Hermes thinks, what tools it can reach, and how it behaves across a session.

On startup, expect to see the active model, available tools, and the current skills context.

Step 4: give Hermes a small real task

Do not start with a benchmark prompt or a giant vague request. A good first task is:

• local
• concrete
• small enough to evaluate quickly

Good examples:

• summarize the structure of a repository
• inspect likely deployment files and configs
• explain what a script directory is doing
• check logs or a local working directory

Hermes makes the most sense when the task is tied to a real environment instead of a generic demo.

What Hermes is doing during the first run

The first run is where Hermes stops feeling like disposable chat software and starts feeling like a runtime.

Officially and practically, Hermes is built around:

• a persistent CLI workflow
• tool use for files, terminal, web, and other capabilities
• reusable skills that can make repeated work more consistent
• continuity across sessions instead of a pure reset-every-time model

That operating model is the real reason many people choose Hermes.

A few commands worth knowing early

You do not need the full CLI reference on day one, but these are useful quickly:

hermes --continue
hermes chat -q "Hello"

Inside Hermes, slash commands are also part of the normal workflow. The official docs highlight commands for help, model changes, new sessions, and browsing skills.

What not to worry about yet

On the first day, you do not need to:

• move straight to Docker
• deploy to VPS immediately
• add a browser frontend before you understand the CLI
• optimize every model and tool setting up front

Start simple first. Add complexity only when you know why you need it.

What to do next

• Read Install Hermes Agent if the environment is not fully ready yet
• Read Deploy if Hermes should move beyond local use
• Read Hermes Agent vs OpenClaw if you are comparing operating models
• Read the official Quickstart if you want the full authoritative walkthrough

Hermes Agent quickstart FAQ

What is the fastest first run for Hermes Agent?

Install Hermes, configure a provider with hermes model or hermes setup, then launch hermes and give it one small local task.

Do I need Docker to get started?

No. For most people, Docker is a later step. Local installation is the clearest and fastest way to understand Hermes first.

Is Hermes Agent a web UI?

No. The official CLI guide is explicit that Hermes is a TUI, not a web UI. Browser-based interfaces can come later through the API server path and tools like Open WebUI.

What should my first task look like?

Keep it small, local, and real. The first run should help you understand the operating model, not test every feature at once.

What if I am coming from OpenClaw?

Hermes already has an official migration path. If that is your situation, compare the two models first on Hermes Agent vs OpenClaw, then use the official migration guide when you are ready.