terminal Hermes Agent
Start
Quickstart · Official docs checked 2026-07-08

Hermes Agent Quickstart

Follow the shortest path from zero to a working Hermes Agent setup. Do not add gateways, cron, skills, or routing until a plain chat works.

Hermes Agent quickstart answer

Use one clean feedback loop before adding automation

  1. Install Hermes Agent with the official desktop installer or CLI installer.
  2. Choose one provider path: hermes setup --portal for Nous Portal, or hermes model for manual provider selection.
  3. Run hermes and confirm a normal prompt gets a useful answer.
  4. Only then add gateway messaging, tools, memory, skills, MCP, cron, voice, or fallback routing.

Before you start

You need

  • • A terminal: Terminal.app, iTerm, Linux shell, WSL2, PowerShell, or Termux
  • • Internet access
  • • One model/provider path
  • • Ten minutes for the first working chat

Not sure which provider?

Use hermes setup --portal first. It gives the simplest guided path through Nous Portal.

1

Install Hermes Agent

Linux / macOS / WSL2 / Android Termux:

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

Native Windows PowerShell:

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

After install, reload your shell: source ~/.bashrc or source ~/.zshrc

✓ Hermes installed successfully

Run `hermes` to start chatting

2

Choose a provider and model

Fast path (guided):

hermes setup --portal

Manual provider path:

hermes model

The provider is the most important setup decision. If model auth or model selection is wrong, Hermes may open but return broken, empty, or unhelpful replies.

3

Verify your first chat

Run:

hermes

Ask:

Say hello and tell me which provider/model you are using.

A successful first chat means:

  • • Hermes responds in normal language
  • • The response is not empty or an auth error
  • • You can ask a follow-up question
  • • The terminal session does not crash
Rule: Do not continue to gateway, skills, cron, or routing until this works.
4

Choose your next path

Goal Do this next Do not do yet
Learn the CLI Read the CLI guide and try file/search/tool examples Do not enable every tool at once
Use Telegram/Discord/Slack Run hermes gateway setup after CLI chat works Do not debug bot tokens before the base provider works
Use local/self-hosted models Verify endpoint URL, model name, and context length Do not add fallback before one model is stable
Resume old work Try hermes sessions list then hermes --continue Do not assume the session is in the current profile
Fix broken setup Run the recovery checklist below Do not keep changing random config values

Troubleshooting

Most Hermes setup failures come from adding too many layers before the first chat works. Start with the base provider and model. Then add gateway, tools, skills, cron, voice, routing, or fallback one at a time.

hermes doctor
hermes model
hermes setup
hermes sessions list
hermes --continue
hermes gateway status
Rule: If the plain chat is broken, do not debug gateway, cron, skills, voice, or fallback yet.

Non-payment success signal

Did Hermes answer your first prompt?

Use this when the install command, provider choice, and first plain CLI chat all worked. It records setup success for GA4/Plausible/Clarity without adding checkout or ad spend.

Quickstart FAQ

Should I run hermes setup or hermes model first? expand_more
Use hermes setup for the guided setup. If you already know your provider and model, use hermes model, save the choice, then verify one normal chat.
When should I configure gateway or messaging? expand_more
Configure Telegram, Discord, Slack, or other gateways only after the plain CLI chat works. If the base provider is broken, gateway setup makes the failure harder to isolate.
What is the fastest recovery checklist? expand_more
Run hermes doctor, confirm the selected provider with hermes model, then start hermes and test a short prompt before changing tools, skills, cron, or fallback.

Quickstart comparison

First-time user

Use hermes setup --portal, run hermes, and stop after one successful chat.

Bring-your-own-key user

Use hermes model, save the provider/model, then test a short prompt before changing tools.

Bot workflow user

Only run hermes gateway setup after the local CLI chat works. This avoids token debugging noise.

Verification before advanced setup

A quickstart is successful only when Hermes can answer in a plain terminal session. Treat install completion, model selection, and gateway setup as separate checkpoints. If you change all three at once, failures become hard to attribute and you may waste time debugging the wrong layer.

After the first useful answer, write down the provider/model that worked, the shell you used, and whether the answer came from the expected profile. Then add one feature at a time: a tool-enabled task, a memory update, a skill, or a gateway message. Re-run the plain chat test after each major change so you always have a known-good baseline.

If you are comparing Hermes with other coding agents, this baseline matters even more. Measure Hermes after it can complete a normal chat, read the right files, and report real command output. That is the point where workflow automation becomes meaningful instead of being hidden by first-run configuration errors.

Source: Hermes Agent Quickstart · Official Docs