Hermes Agent, in its own sandbox
Run Hermes Agent, Nous
Research’s open source agent, always on, in a sandbox of its own: a
workspace hermes, which is a separate account on your server with its own
files and model keys, no sudo, and no way into your other workspaces.
You need: nothing beyond Getting started.
Before you start: machine, skills, workspace
curl -fsSL https://mydevmachine.sh/install.sh | sh
devmachine setup
devmachine skills add
devmachine workspaces new hermes
devmachine sync
Already have a machine? Skip setup. Already have the workspace? Skip the
last two. See getting started for what each command
does.
By hand
1. Install Hermes Agent
devmachine ssh hermes
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
The installer sets up Python, Node and the other tools Hermes needs itself, so the workspace does not need them added first.
2. Add a model
hermes setup --portal
This walks through picking a model. It supports the Nous Portal (one login
covers a model plus its tool set), or your own OpenAI or OpenRouter key —
hermes model switches later, and hermes config set stores a key by
hand. Whichever you pick, the choice and the key stay in this workspace’s
own account, never in your configuration.
3. Start it, and keep it running
hermes
You do not need tmux yourself: every SSH login to a workspace already lands
in a tmux session called main (the zsh package does this), and it stays
up after you log out. Detach with Ctrl-b d, or just close the terminal —
devmachine ssh hermes puts you back in the same session.
For messaging apps (Telegram, Discord, Slack, WhatsApp, Signal) instead of
the terminal, run hermes gateway setup then hermes gateway start. To
make Hermes survive a reboot too, not just a closed terminal, see the
project’s own docs on daemon and background backends.
With your agent
Open a session on your own computer (devmachine skills add taught it the
CLI) and say:
Create a devmachine workspace called hermes and install Hermes Agent
in it, following the project's own install script.
The agent creates the workspace, SSHes in, and runs the installer. Picking a
model and running hermes setup --portal (or handing over an API key) is
yours to do — a login or a key is not something the agent can do for you.
Check it: devmachine ssh hermes again, then hermes — it comes back
to a working session with the model you configured.
Source: Hermes Agent, Hermes Agent on GitHub
Where to go next
- Agents
OpenClaw, in its own sandbox
Run the OpenClaw agent around the clock in a workspace of its own, away from your laptop.
Beginner
- Agents
Pi as your coding agent
Install the Pi coding agent in a workspace and use it for everyday work on your server.
Beginner
- Agents
Agno AgentOS with a control plane
Serve Agno AgentOS and its control plane dashboard, each at its own subdomain with HTTPS.
Advanced