Run a Hermes chat agent
Your own AI agent in Telegram, Discord or Slack, always on, with its memory and skills kept on App storage.
Hermes Agent by Nous Research is an AI agent that runs on a server, answers in Telegram, Discord, Slack and other chat apps, and keeps its memory and skills between conversations. The hermes-agent-starter template runs it on Dockhold with an HTTPS address, its data on App storage and its admin password in Secrets. You set it up from your browser. Want Hermes to work on your code from GitHub issues instead? That is the Hermes coding agent.
1. Open the deploy link
Open the deploy link and sign in if asked.
2. Size it and add App storage
Under App size, pick 1 GB. Under App storage, turn it on and pick 10 GB. This app needs a paid plan: its build is larger than the free plan allows, and Hermes uses about half a gigabyte of memory before its first message. See pricing.
3. Add the admin password
Under Environment, in the Secrets list,
click New secret. Give it a name that belongs to this
app, for example hermes-agent-admin-password, and a password
of at least 12 characters. Tick it and set its
Env var name to ADMIN_PASSWORD. Secrets are
shared across your apps by name, so a name of its own keeps this
password separate from your other apps.
4. Deploy
Click Deploy and wait until the app shows as running.
5. Sign in and pick a model
Open the app's URL and sign in as admin with that password.
The setup page opens. Pick a model provider, paste its API key and enter
a model (OpenRouter has free
models). Tick Telegram, paste the bot token from
@BotFather, and click
Save & Start.
The key is yours. You pay the model provider directly, and Dockhold never pays for model use or resells it.
6. Approve yourself
Send your bot a message. It replies with a pairing code, and the request appears under Users in the admin panel. Click Approve, and the bot answers you from then on. Anyone else who messages it gets a code too, and waits for your approval.
What you get
- A setup page at
/setup: pick a provider and model, tick chat channels and tools, no config files to edit. - The Hermes dashboard at
/, behind the same login, with chat, keys, skills and a live log. - Config, pairings, chat history, skills and memory on App storage, kept across restarts and deploys.
- Backup and restore: download a zip of the agent's config, chat history and skills, and restore it later.
The setup page lists OpenRouter, Anthropic, Google AI Studio and many other providers, plus any OpenAI-compatible endpoint. Chat apps: Telegram, Discord, Slack, WhatsApp, Email, Mattermost and Matrix.
Upgrading
The app stays on the version it was built with. Restart starts the same build again. Build again on the app page rebuilds from the starter's latest commit, which can move Hermes to a newer release. App storage is not backed up, so download a backup from the admin panel first. The Update button inside the Hermes dashboard does nothing here on purpose.
To choose when you upgrade, click Use this template on GitHub and deploy your own copy. Every push to it then builds and starts the new version.
Troubleshooting
- If the app refuses to start, its logs on the app's Activity tab show one line saying what is missing. App storage is on the app's Size tab. Secrets are attached on its Variables tab, and their values are edited under Secrets in the dashboard sidebar. Fix it and click Restart.
- Free models are often rate-limited by their provider. If the bot goes quiet, check the Logs panel in the admin panel, and pick another model in the setup page if you see a rate-limit error.
The full reference is in the template's README.
Credits
Hermes Agent is made by Nous Research. The admin panel and setup page come from the Hermes Agent template by praveen-ks-2001.