Set up the Hermes Agent

Hermes is a self-hosted AI agent by Nous Research (GitHub repo). With lowcloud Platform you deploy your own Hermes instance into your project with a single click. No server provisioning, no container setup, no manual volume management.

Hermes does not get a public URL by default. Your instance stays private until you change that yourself. Setup happens entirely in the browser via the service's Shell tab.

Prerequisites

• A project on lowcloud Platform
• An API key for your model provider (e.g. Anthropic)

Deploy Hermes

In your project, click + Create Service → Hermes Agent.

1. Name your service (e.g. hermes). Other services in this project will reach the instance under this name later.
2. Pick a container size tier. Hermes is an AI agent and needs resources accordingly. The minimum is Medium, we recommend Large.
3. Click Deploy Hermes Agent.

A 10 GiB block storage volume is provisioned automatically and mounted at /opt/data. That is where Hermes stores its configuration and state, both of which survive restarts and redeploys.

Onboarding via the Shell

After the deploy, the service starts without any configuration and stays stable in Running state. Before you can use Hermes, run through the onboarding once:

1. Open your Hermes service and switch to the Shell tab.
2. Run the setup:

hermes setup

3. Follow the interactive wizard. This is where you connect your model provider (e.g. with an API key) and configure the agent. Ideally, connect a messaging channel right away so you can reach your agent conveniently from your chat app afterwards.

The configuration is stored on the persistent volume at /opt/data, so you only need to run the setup once.

Even without a messaging channel, you can always interact with Hermes via the shell:

hermes

Optional: enable the API

Hermes ships with an OpenAI-compatible API server. It is off by default and the service does not bind a port. To enable it, set three variables in the Environment tab:

New values take effect after the next redeploy. The API server then listens on port 8642.

Other services in the same project reach the API over the internal network using the service name:

http://<service-name>:8642/v1

Authenticate with the key as a bearer token:

curl http://<service-name>:8642/v1/models \
  -H "Authorization: Bearer <API_SERVER_KEY>"

Summary

• + Create Service → Hermes Agent, pick a name and tier, deploy.
• 10 GiB of persistent storage at /opt/data is included automatically.
• Open the Shell tab once, run hermes setup, and ideally connect a messaging channel along the way.
• Optional: enable the API server via environment variables and use it internally at http://<service-name>:8642/v1.

For more details on Hermes itself, see the official Hermes docs and the GitHub repo.