17 min read
A step-by-step walkthrough for setting up Hermes Agent, the open-source AI agent from Nous Research.
TL;DR
Tap to expand the short version
- Hermes Agent is an open-source AI that runs 24/7 on a server, not just in a chat window on your machine.
- This guide walks the non-technical path, rent a $7 to 10/month VPS, deploy in one click, add your API key, and you’re running. No terminal required.
- Part 1 covers core setup. Part 2 (coming later) adds scheduling, phone messaging, and real business tool connections.
- Total cost, about $7 to 10/month for hosting plus whatever you spend on your AI model provider. Setup takes 30 to 45 minutes.
- You don’t need to code, but you do need a credit card and a functioning API key from OpenAI, Anthropic, or another provider.
In this guide
Why Hermes Agent
A standard chatbot is a temp worker who wipes their memory at 5 PM. You close the tab, and the context dies.
Hermes stays running, retains selected context and can execute scheduled tasks. Maintained by Nous Research, it is closer to rented infrastructure than software you open for a single conversation.
Most of the software we use lives locally on a device or inside a browser tab. Hermes requires a physical space that does not power down when you do. For anyone using it as a genuine business assistant rather than a weekend toy, that means provisioning a small, always-on server.
The Infrastructure Requirement
Hermes will run while you sleep, but only if you pay for the desk.
To stay available when your laptop is closed, Hermes must run on a virtual private server (VPS). rented capacity in a data centre that remains powered and connected. Install Hermes there and it can continue running independently of your computer.
Once deployed, the agent stays alive as long as three conditions are met. The server bill is paid. The dashboard connection is configured. The AI model provider account has funds.
If a prepaid model balance runs out or a credit card fails, the agent does not crash or throw a fatal error. It simply stops answering. You will only notice when you check in and find silence. Check the provider balance occasionally once real workflows depend on the agent.
Memory and skills
The setup screens use specific engineering terminology. You do not need to understand the underlying code, but you do need to understand the operational difference between the terms.
The agent runs inside a container. This is a sealed environment that keeps everything the agent touches isolated. If the agent makes a mistake or runs a bad command, the container prevents the error from reaching the host machine.
Your API key is the string that allows Hermes to talk to your AI model provider. Functionally, it is a password. It belongs in one specific field in the dashboard environment settings. It never belongs in a chat message.
Its architecture separates knowledge into two categories.
Tools are individual actions. Searching the web, reading a file, or browsing a page. One tool equals one action.
Skills are procedures. When the agent figures out a complex way to combine tools into a finished job, it saves that approach to reuse later. The documentation calls this procedural memory. In business terms, Skills are the Standard Operating Procedures.
Memory consists of small, durable facts about you and your setup. In business terms, Memory is the CRM.
The two exist apart because facts and procedures serve different purposes. Memory is what the agent knows about your company. Skills are what the agent knows how to do with that knowledge.
Two deployment paths
There is no single correct way to get Hermes running. There are two paths, and they solve entirely different problems.
Option A. The Local Install (The Official Way)
This is the fastest route and the method recommended by the developers. It is ideal for testing the software. You download the desktop installer from the official Hermes website and run it. No command line is required. Alternatively, Mac, Linux, and WSL2 users can install it via a single terminal command. Windows users can run the equivalent PowerShell script.
Once installed, you run hermes model to select your AI provider, or hermes setup --portal to use a unified subscription. Type hermes to launch the chat.
This version only runs while your computer is awake. If you close the laptop, every scheduled job stops. If you just want to see what the agent feels like, stop here. If you want it to work the night shift, you need Option B.
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
Option B. Run Hermes on a server
Choose the server route if Hermes must run scheduled tasks, remain reachable overnight or connect to your phone without relying on your laptop. The rest of this guide uses a hosting provider’s one-click deployment, so the setup does not require terminal work.
Permissions and security
Hermes can run real commands, and the protection comes from the operating-system-level box those commands run inside. This is why agents that bypass containment or run outside a container represent a real risk. Learn more about false confirmations.
The one-click Docker deployment provides container isolation by default. Hermes also checks commands against known dangerous patterns. In manual mode, every flagged command waits for approval; smart mode uses a secondary model and escalates uncertain cases; off disables the checks and is unsuitable for a first deployment.
Container deployment limits the damage from a mistaken command by keeping it inside a box that can be rebuilt. It does not replace command approval. Keep both controls enabled.
The practical rule is to keep Hermes inside its container and leave command approval enabled. Do not run it as a bare process on your everyday computer merely to avoid an approval prompt.
Suitable work
The setup in this guide gets you a general-purpose agent, not a sales agent or a finance agent. The steps are identical regardless of which part of the business you have in mind. Nothing in Part 1 is sales-specific or developer-specific. What differs, later, is which skills you teach it and which tools you connect it to.
| Function | Possible right after Part 1 | Needs Part 2 or later |
|---|---|---|
| Sales | Draft a follow-up email or call script when you ask, in conversation | Draft follow-ups on a schedule, or pull context from a connected CRM automatically |
| Marketing | Brainstorm campaign angles, summarise research you paste in | Check competitor pages daily and message you what changed, unprompted |
| Finance | Answer general questions, run calculations in conversation | Log expenses from a forwarded receipt, produce a recurring weekly summary |
| Social media | Draft post copy on request | Post on a schedule, monitor mentions, let you approve drafts from your phone |
| General ops and admin | Answer questions with real memory of your business context | Morning briefings, recurring reminders, breaking a task into subtasks automatically |
Who this guide suits
This guide is for business and operations readers who want to understand the setup before giving Hermes real work. Readers comfortable with a terminal will move faster with the official quickstart.
What You Need Before You Start
- An account with any AI model provider like OpenAI, Anthropic, or an aggregator such as Nous Portal or OpenRouter
- 30 to 45 minutes without interruption for the first setup
- A password manager, or at minimum a private note, to store the admin credentials you’re about to create
- Your credit card for a small monthly server charge
Upfront Costs
Two bills determine the cost. The server is fixed; model usage varies with workload. Treating them as one figure is a common source of surprise.
| Item | Typical cost | Notes |
|---|---|---|
| VPS (KVM 2. 2 cores, 8GB RAM, 100GB NVMe storage) | Roughly $6.49 to $9.99/month on a 12 to 24 month promotional term | Renewal rates run noticeably higher than the intro rate; check the live price at checkout before committing to a term length |
| AI model usage, light daily conversation only | Low, typically single-digit dollars monthly on a moderately priced model | Scales with message volume, not with time the server is on |
| AI model usage, with frequent scheduled background jobs | Can exceed the server cost | Each cron job and each memory-compression pass consumes tokens even when you never open the chat window yourself |
| The Hermes software itself | $0 | Open source, MIT licensed, no subscription for the agent framework |
A light setup may cost little beyond hosting. Several scheduled jobs using a flagship model can make inference the larger bill. There is no reliable flat figure without knowing the workload.
Step 1. Choose Your Model Provider
Hermes provides the agent system. The connected model provides the intelligence and can be changed later without rebuilding the deployment. Three starting options cover almost everyone.
You need an API provider. OpenAI and Anthropic offer direct access; Nous Portal bundles models with services such as search and speech; OpenRouter provides one key for models from several labs. Choose based on the models and billing arrangement you need.
Before Step 3, you need either an API key or a Nous Portal login that you can authorise.
A ChatGPT Plus or Claude Pro subscription does not include API access. Generate a separate key through the provider’s developer platform unless you already have one.
Copy a new API key immediately because most providers show it only once. If you lose it, generate another.
Step 2. Rent Your Server
A VPS is rented computing capacity in a data centre. The provider maintains the hardware and connection while you control the software running on it.
This guide uses Hostinger because its VPS catalogue provides a one-click Hermes deployment and removes manual Docker configuration. Hermes can also run on other Docker-capable VPS providers.
Go to the VPS section of the hosting dashboard and select the KVM 2 plan. 2 CPU cores, 8GB RAM, 100GB NVMe storage. That specification is sufficient for a single agent running a handful of background jobs; a smaller plan will work for pure conversation but tends to strain once you add scheduled tasks in Part 2.
Choose a billing term deliberately. shorter contracts cost more per month but reduce commitment. Select a data centre close to the main user, since location affects latency more than the billing term.
For a first server, choose the data centre closest to the main user. Migration can wait until usage proves it is needed.
VPS checkout pages often default to the longest contract. Choose a shorter term if you want to test Hermes before making a long commitment.
Step 3. Deploy the Hermes Container
Once the server is provisioned, you’ll land in its management dashboard. From here the path is, Docker Manager, then Projects, then Compose, then the one-click deployment catalog. Type Hermes into the catalog search.
The catalogue contains three similar options. Hermes Agent is the minimal core. Hermes Workspace adds a dashboard, persistent memory, a skills marketplace and a terminal. Hermes Web UI is only a front end and requires an Agent deployment underneath it.
For this guide, deploy Hermes Workspace. It adds a dashboard, persistent memory, a skills marketplace and a terminal. Plain Hermes Agent has fewer moving parts and suits readers who prefer simplicity. Avoid Hermes Web UI unless an Agent deployment is already running.
The configuration window asks for a workspace password and, optionally, the API key from Step 1. Save the password in a password manager. You may leave the API field blank and add the credential later through the dashboard’s dedicated Keys tab; do not paste credentials into chat.
Click deploy. The container should build within a few minutes and then show as running in Docker Manager. If progress appears stalled, refresh once before troubleshooting.
Step 4. Connect Your Model
From the Docker Manager project row, click Open. This launches the Hermes dashboard in your browser. Log in with the workspace password you just set.
Go to Models in the left-hand navigation. If you left the API key blank in the previous step, paste it into the Keys tab now. Never type it into a chat message to the agent.
Once connected, set your main model for everyday conversation. Choose a balanced, moderately priced model. You can change this in two clicks later, so there is no need to default to the newest or most expensive option.
Leave the auxiliary tasks setting on automatic. This governs background housekeeping, like compressing long conversations or summarizing shared documents. The automatic setting routes this invisible work to a cheaper model. It keeps your main model’s token costs predictable.
You will know this step worked when the Models screen shows your chosen model as active. A connection error almost always means the key was mistyped or pasted with a stray space. Re-copy it directly from the provider’s dashboard. Do not retype it by hand.
Step 5. Send Your First Message
Open a new chat session and type a simple greeting. A reply confirms the entire chain is working. The server is running. The container is healthy. The model is connected.
This is also the moment to feed the agent two or three true, specific sentences about your business. From this point forward, it starts keeping notes. That persistent memory is the entire difference between this infrastructure and a standard chat window.
Part 1 ends with an agent that runs independently of your laptop, responds through the dashboard and remembers selected information. Messaging, scheduled tasks and deeper business-tool connections come in Part 2.
Operational Risks and Common Failures
Disabling command approval to move faster. The approval system exists to catch destructive commands before they run. Disabling it removes a critical safety check. It is not a nag screen to bypass for speed.
Running Hermes outside its container. The container is your main layer of protection. It contains a mistaken command to a box you can wipe and rebuild. Running the agent bare exposes your host machine directly. It is a strong layer, but not absolute. Mounted volumes and network access can still create exposure. The container must work alongside the approval system, not replace it.
Pasting an API key into the chat window. Keys belong in the dashboard’s dedicated keys tab, written to the server’s environment file. Chat content gets logged, summarized, or shared with external tools the agent calls. Never feed credentials into the conversation.
Losing the workspace password. There is no shortcut back into the dashboard without it. Write it down correctly the first time.
Next steps
Right now, the agent can talk and remember. It cannot reach you outside the dashboard, run unsupervised schedules, or touch your business tools.
Part 2 will connect Hermes to a phone messaging service, add a scheduled task, define memory and skill boundaries, and configure one recurring business workflow.
For the interface between an agent and its tools, read the Code Mode explainer. The IC Lab repository guide groups other implementation projects by job. More system explainers sit in Under the hood.
FAQ
Is Hermes Agent Completely Free to Use?
The software itself is free and open source under an MIT license. However, you will incur separate costs for the server hosting the agent and the usage fees associated with the AI model you connect to it.
No, you do not need to know how to code. You can use a one-click Docker deployment through a hosting provider’s dashboard to avoid using the terminal entirely.
If you deploy your agent to a VPS, it will remain online independently of your personal devices. The desktop app version, however, will only run while the application is actively open on your machine.
There is no single required model, so a balanced and moderately priced option is a great starting point. You can easily switch models later without losing your existing configuration, memory, or skills.
No, your initial choice is not a permanent commitment. Hermes Agent allows you to switch between different AI models with just a few clicks whenever you choose to update your setup.
Memory holds small, durable facts about you and your setup that should always be available. Skills hold longer procedures for finishing a specific kind of task, saved once the agent has worked out how to do it, and loaded only when a job calls for them. The project describes memory as what the agent knows and skills as what it knows how to do.
Yes. Hosting providers generally let you upgrade a VPS plan without a full reinstall, and Hermes itself doesn’t require a specific plan tier to function, only enough headroom for how many background jobs you’re running at once.
Yes, as long as it’s deployed to a VPS rather than the desktop app, and as long as the server is paid up and your model provider account has funds or an active subscription. The one silent failure mode is a depleted model balance, the agent doesn’t crash, it just stops replying, so check your provider’s balance occasionally once you’re relying on it.
No. Every business function starts from the identical deployment covered in this guide. What differs is what you build on top afterward, which skills you teach it and which tools or platforms you connect, and that starts in Part 2, not here.
ChatGPT is a conversation you initiate. Hermes is infrastructure that stays running, holds tools and scheduled jobs, and reaches you across whichever messaging platforms you connect it to, using whichever underlying model you choose, including OpenAI’s models if you want them. It’s a different category of product built on top of the same kind of model, not a competing chatbot.
Reasonably, if you follow the basics. keep the approval system on rather than disabling it, keep credentials in the dashboard’s keys tab rather than in chat, and rely on the container as your main containment layer, not an absolute guarantee. A command the agent shouldn’t have run generally stays confined to a box you can wipe and rebuild, though misconfiguration, mounted volumes, or exposed network access can weaken that isolation, which is why the approval system stays on rather than being treated as optional.
Spotted an error? Contact the desk. Our corrections policy explains how we handle corrections and updates.
For other projects to run on your hardware, browse IC Lab’s GitHub repo directory and compare setup requirements and licences.