How to Use Hermes Agent: Install, Set Up, and Run Your First Job
A practical, step-by-step walkthrough of getting Nous Research's autonomous agent running and doing useful work.
Hermes Agent is quick to install, but the first run has a few choices that trip people up. This guide walks you through each one in order.
You will install Hermes Agent, point it at an AI model, connect a chat app, and schedule your first job. Every step uses the official commands from Nous Research.
By the end, you will have a working agent that runs on its own and reports back to you. No prior agent experience is needed, though basic comfort with a terminal helps.
What you need before you start
Before you install Hermes Agent, you need a machine to run it on and an AI model to power it. The agent is the worker, but the model is its brain.
You can run it on your own computer, a small cloud server (VPS), or the managed Hermes Agent Cloud. For a first test, your own laptop is fine.
You also need an AI model provider. That can be a paid API key from a service like OpenRouter or OpenAI, or a free local model you run yourself.
- A host — your laptop, a Linux VPS, or Hermes Agent Cloud.
- A model provider — Nous Portal, OpenRouter, OpenAI, a custom endpoint, or a local model.
- A few dependencies — Python 3.11, Node.js, ripgrep, ffmpeg, and Git (the installer handles most of this).
- A chat app (optional) — Telegram, Slack, or Discord if you want mobile replies.
Want your first Hermes Agent job, like a daily briefing or a monitoring alert, live this week? We set up and govern autonomous agents for SMBs. Book a consultation to scope it with our team.
Book a ConsultationHow do you install Hermes Agent?
You install Hermes Agent with a single command from the official Nous Research installer. It works on Linux, macOS, WSL2, Termux, and native Windows.
On Linux, macOS, or WSL2, open a terminal and run the install script. Then reload your shell and start the agent.
On Windows, use the PowerShell command instead. The installer bundles a small copy of Git if you do not already have it.
- Linux / macOS / WSL2:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash - Then:
source ~/.bashrcand start it withhermes - Windows (PowerShell):
iex (irm https://hermes-agent.nousresearch.com/install.ps1) - Prefer containers? A Dockerfile and docker-compose file ship with the project, so you can run it in Docker instead.
How do you connect an AI model?
You connect a model with the hermes model command, which lets you pick a provider and paste an API key. This is the step that gives your agent its intelligence.
The fastest path is hermes setup --portal, which logs you into Nous Portal over OAuth and gives unified access to many models. You can switch providers later.
Hermes Agent is model-neutral, so you are not locked in. You can point it at OpenAI, OpenRouter, a custom endpoint, or a local model through Ollama for full privacy.
What happens on your first run?
On first run, Hermes Agent starts an interactive session where you type goals in plain language and watch it work. Start with something small and safe.
Give it a read-only task first, like summarizing a web page or a local file. This lets you see how it plans and acts before you grant riskier access.
The agent remembers this session. Its persistent memory means the next time you ask a related task, it already has context and will not ask you the same questions twice.
How do you use skills and memory?
Hermes Agent builds skills on its own after it finishes a complex task, saving the know-how so it can repeat the job faster next time. You do not have to write them by hand.
Skills are stored locally and follow the open agentskills.io format, so they are portable and shareable. You can also install skills others have made.
Memory works alongside skills. The agent keeps a searchable record of past sessions, so a weekly report it ran last Friday informs the one it runs today.
- Automatic skills — the agent writes a reusable skill after hard tasks.
- Portable format — skills follow the agentskills.io open standard.
- Persistent memory — past sessions are searched for context on every run.
- Local by default — when you self-host, this data can stay on your machine.
How do you schedule your first unattended job?
You schedule a job by describing it in plain language and setting when it should run, using the built-in cron scheduler. This is what turns Hermes Agent into a background worker.
A good first scheduled job is a morning briefing. Ask it to summarize your calendar and a news source each day at 7 a.m. and send the result to your chat app.
This is the same set-and-forget pattern we use at Layer3Labs to run production reports every day. Once the schedule is set, the agent runs with no one watching and delivers the result to you.
Common Hermes Agent commands
A handful of commands cover most of what you will do day to day. Keep them close while you learn.
These map to the questions people ask most: how to start it, change the model, connect a chat app, and update it.
- Start the agent:
hermes - Set up the model:
hermes model - Portal login:
hermes setup --portal - Connect a chat app:
hermes gateway setupthenhermes gateway start - Update: re-run the install command to pull the latest version.
Should you self-host or use Hermes Agent Cloud?
Choose self-hosting for control and privacy, or Hermes Agent Cloud for a hands-off setup with no server to maintain. Both run the same agent.
Self-hosting keeps your data and keys on hardware you own, which matters for regulated or sensitive work. You trade that for patching and uptime duties.
Hermes Agent Cloud uses serverless persistence, so your agent hibernates when idle and wakes on demand to keep costs low. It is the faster route for non-technical teams.
| Factor | Self-host | Hermes Agent Cloud |
|---|---|---|
| Setup effort | Medium | Low |
| Data control | Full, on your hardware | Managed for you |
| Maintenance | You patch and monitor | Handled by the host |
| Best for | Privacy-sensitive, technical teams | Fast start, non-technical teams |
Frequently Asked Questions
- You install Hermes Agent with one command from the official Nous Research installer. On Linux, macOS, or WSL2, run the curl install script, reload your shell, then start it with the hermes command. Windows uses a PowerShell one-liner instead.
- You do not need to code to use Hermes Agent, but basic terminal comfort helps for setup. Once it is running, you give it tasks in plain language and it replies on chat apps. Non-technical teams often use the managed cloud option or an implementation partner.
- You connect a model with the hermes model command, or run hermes setup --portal to log in through Nous Portal. Hermes Agent is model-neutral, so you can use OpenAI, OpenRouter, a custom endpoint, or a local model.
- Yes. Hermes Agent runs on native Windows using a PowerShell install command, and also on WSL2 if you prefer a Linux-style setup. The installer bundles a small copy of Git when one is not already present.
- You use the built-in cron scheduler and describe the job in plain language, setting when it should run. The agent then runs unattended and delivers results to a chat app or email. A daily morning briefing is a good first scheduled job.
- Yes. Hermes Agent is model-neutral and can point at a local model through Ollama or a custom endpoint. Running a local model keeps your data on your own hardware, which is useful for private or regulated work.
- You update Hermes Agent by re-running the official install command, which pulls the latest version. Because it is open-source under the MIT license, you can also update by pulling the newest code from the GitHub repository.
- Start with read-heavy, low-risk jobs that are easy to check, like daily briefings, monitoring, or research summaries. Prove the agent is reliable before you grant access to money, customers, or production systems.
Want Hermes Agent set up and running safely?
We install, connect, and govern autonomous agents like Hermes Agent for SMBs every day. Book a consultation and we will get your first job live without the risk.
Book a Consultation