Set up Hermes Agent on your server

Hermes Agent is a self-improving AI agent by Nous Research. It keeps memory between sessions and gets better at the work you give it. It runs on your server with the AI model you choose, and installs free on every plan.

Install it

1. Picked Hermes Agent at checkout? It installs by itself as soon as your server is ready. Open Servers and pick your server: the top of its Overview shows the install and a Get started list with the next steps.
2. Otherwise, on your server's Overview press Install (free) on the Hermes Agent card (on a server you set up as just a server, press Want an app? Install one free first). The server must be running. You then land on Hermes Agent's tab, which shows the progress. The first app on a server also prepares it, so the progress says "Preparing your server" for a minute before the install starts. The whole install takes a few minutes.
3. Hermes Agent starts by itself when the install is done.

Connect a model

Hermes Agent needs an AI model to think with. Once it is installed, press Connect AI in the Get started list on your server's Overview, or open Hermes Agent's tab in the server menu and choose AI model:

• Paste your ScaleMax key (on a phone, press Paste beside the box). It is checked by itself (or press Check key), the page shows how many tokens are left and picks a model, and you press Connect. Choose a different model lists every model the key offers.
• No key yet? Get a key on ScaleMax opens ScaleMax, where one key covers many models on one balance. Buy one, copy it, then come back and press Paste.
• Use another AI provider: choose OpenAI, Anthropic, OpenRouter or a custom address, paste your key, type the model name exactly as your provider lists it and press Connect. A custom address must be a public https address, OpenAI- or Anthropic-style.

ScaleMax, OpenAI and OpenRouter keys, and keys for an OpenAI-compatible custom URL, are checked with the provider first. Anthropic keys, and keys for an Anthropic-compatible custom URL, are not checked: a wrong one shows up when Hermes Agent first tries to answer. The key is then written straight into Hermes Agent's settings inside your server. Your dashboard keeps only the provider, the model and the key's last four characters, shown as "Connected: provider · model · key ending …1234". Hermes Agent restarts with the new key in a minute or two.

With a model connected, Change model switches to another model your key offers, keeping the same key. It takes effect in a few seconds, without a restart. There is no limit on how often you check a key or change the model.

If your provider's website protection (for example Cloudflare) blocks requests from servers, the dashboard says so instead of calling your key wrong. Ask the provider to allow API access from servers, or use another provider.

If Hermes Agent stops replying because the key has run out of balance, is refused, can no longer use its model or is blocked by the provider's firewall, we tell you: a notice under the bell in your dashboard, and an email unless you turn off AI app alerts in Account → Email preferences.

Open it

The running app opens in its own tab in the server menu, already signed in. Use Open in new tab for a full window, or Reload if the page stops answering. Hermes Agent lives at your server's app address:

https://vmN-app.box.scalemaxcloud.com

Swap vmN for your server's name. The address sits behind your ScaleMax Cloud login, so open it from the dashboard. Anyone else gets the sign-in page.

Hermes Agent's own dashboard login is handled for you; there is no separate password to keep.

Chat on Telegram and WhatsApp

Talk to Hermes Agent from Telegram or WhatsApp. Set it up in Hermes Agent's tab in the server menu, under Chat apps, which shows while Hermes Agent is the active app. No terminal needed. Only you can chat with it: messages from anyone else are ignored and never reach your AI model, so nobody else can spend your credits. Group chats are off.

Telegram. In Telegram, open @BotFather, send /newbot and answer its two questions. It replies with a token. Paste the token, send your new bot any message, then press Find my Telegram ID and pick yourself (or type your numeric ID; @userinfobot tells you what it is). Press Connect Telegram. The token goes straight into Hermes Agent on your server; your dashboard keeps only its last four characters.

WhatsApp. Choose a separate number for the agent (recommended) or your own number, enter your WhatsApp number with the country code, and press Link WhatsApp. The first link adds WhatsApp support to Hermes Agent, which takes a minute or two. Then a code appears on the page: on the phone with the agent's WhatsApp, open Settings → Linked devices → Link a device and scan it. The page turns to Linked by itself. With your own number, chat with the agent in "Message yourself".

Disconnect and Unlink remove a chat app from Hermes Agent. After unlinking WhatsApp, also remove the device on your phone under Linked devices. Check status asks Hermes Agentwhether each chat app is connected right now.

• Chat apps answer only while Hermes Agent is running. Stop pauses them and Start Hermes brings them back. Replacing Hermes Agent with another app deletes its Telegram and WhatsApp links too.
• WhatsApp links work like WhatsApp Web, so the phone must come online now and then or WhatsApp unlinks it.
• Each server allows ten Telegram checks, connects and WhatsApp links an hour, counted together.

Switching apps

Hermes Agent keeps its memory, skills and settings in /opt/hermes-data on your server.

One app runs on a server at a time. To change it, press Switch app on your server's Overview, or Switch to another app under Settings in Hermes's tab, and pick the new app. You then choose one of two things:

• Replace Hermes with the new app: Hermes Agent and all its data (chats, memories, settings, connected Telegram and WhatsApp, and the saved AI key) are permanently deleted from this server, then the new app installs, in about 3 to 5 minutes. You type Hermes to confirm first, and it can't be undone.
• Keep Hermes and set up the other app yourself, with its guide. The dashboard adds no tab for that app and never touches it.

A server set up before one app per server can have several apps installed. There, switching to one that is already installed deletes nothing: Hermes Agent is paused with its data kept, and you can switch back. Replacing them with an app that is not installed deletes all of them, and you type delete to confirm. Stop pauses Hermes Agent too and keeps its data; Start Hermes runs it again.

Limits

• One app runs at a time on a server.
• Hermes Agent is light: it idles around 239 MB of RAM, so KVM 1 (1 vCPU / 4 GB) is enough for API-only use. KVM 2 (2 vCPU / 8 GB) leaves room for heavier tools.
• Model usage is billed by your model provider, or from your ScaleMax balance with a ScaleMax key.