Self-hosting Hermes requires provisioning a VPS, installing Docker, configuring Caddy or Nginx, managing environment variables, and maintaining everything when it breaks. That is a lot of work before the agent does anything for you.
Hivra takes all of that away. The entire infrastructure layer, including server provisioning, container runtime, networking, SSL and monitoring, is handled before you sign in.
What self-hosting actually requires
To run Hermes Agent on your own VPS, you need a server running Linux (Ubuntu 22.04 or 24.04 is recommended), Docker and Docker Compose installed and configured, a domain name with DNS pointed at the server, a reverse proxy (Caddy or Nginx) configured for SSL termination, and environment variables correctly set across a ~30-variable `.env` file.
You also need a backup strategy for the persistent memory volume, monitoring so you know when the container crashes, and a restart policy so the agent comes back up after reboots. None of this is extraordinarily difficult if you know Linux, but it takes real time up front, and the work is ongoing. Each major Hermes update can require migration steps, dependency updates, or config changes.
When it breaks at midnight because the Docker daemon failed or the SSL certificate did not auto-renew, you are the one who fixes it.
What Hivra handles instead
Server provisioning is automatic. When you launch an agent, Hivra creates its server and starts the Hermes Agent container, configured and tested for stability. You do not select an OS, configure SSH keys, or install anything.
Networking and SSL are handled. Your agent's web interface is served over HTTPS on a Hivra address, with no Nginx config files.
Monitoring runs in the background. If the agent process crashes, it restarts automatically. Recovery from a host failure depends on the agent and provider and is not guaranteed.
Updates are managed. When Nous Research ships a new Hermes Agent version, we test it on one agent before rolling it out to everyone. Your memory and configuration carry forward.
The sign-up to running agent flow
Create an account and choose a plan. Infrastructure provisions automatically, and a status indicator shows progress while your agent comes up.
Go to the Keys section and paste your AI provider API key (Anthropic, OpenAI, or an OpenRouter key). This is the only credential you need to configure.
Optionally set up your agent's system prompt to tell it who you are and what you want it to do. This can be done at any time, and the agent runs with sensible defaults immediately.
That is the full setup. The agent is live and accessible from the dashboard. Scheduled tasks, integrations, and additional configuration happen in the UI without touching a config file.
For users who do want terminal access
If you want to inspect the container, add custom tools, or configure something not exposed in the UI, the dashboard exposes log viewing, the agent's tool directory, and a container shell, without managing your own SSH keys or sshd.
This is opt-in. You do not need it to use the platform, and enabling it does not change anything about the standard managed behavior. It is there for users who want the managed hosting baseline but also want escape hatches to the infrastructure.
What's included
- No VPS to provision or maintain
- No Docker, Compose, or container management
- No Nginx, Caddy, or reverse proxy configuration
- No SSL certificate management
- Automatic container restarts and health monitoring
- Hermes updates tested before rollout
- Status indicator while your agent's server comes up
- Dashboard escape hatches: log viewer, tool directory, container shell
Common questions
Do I need any technical knowledge to use Hivra?
Basic familiarity with AI tools is enough. You do not need Docker, Linux, or networking knowledge. The sign-up flow walks you through everything.
What if I want to customize the underlying container config?
Advanced users can access container configuration, the agent's tool directory, and a container shell through the dashboard. You keep access to all of it, and we do not require it.
Is Hivra just a VPS with a UI wrapper?
No. Hivra is managed infrastructure for AI agents, including Hermes, OpenClaw, Claude Code and Codex: a pre-configured computer for each agent, pre-installed browser automation, managed updates, and a dashboard built for agents. A generic VPS does none of this for you.
What region is my server in?
Hivra chooses where hosted agents run; there is no region picker at signup today. Hermes runs on Hivra Cloud only. Other agents, such as OpenClaw, Claude Code and Codex, can run on your own Hetzner Cloud project (in preview), where you pick the location.
What happens if Hermes releases an update that breaks my configuration?
We test Hermes updates on one agent before rolling them out to everyone. If an update still breaks your setup, contact support and we will help you fix it.