Hermes Agent Backend: Terminal Backends and hermes serve Explained

“Hermes Agent backend” means two things. First, the terminal backend — where Hermes actually runs commands, from your own machine to Docker, SSH or a cloud sandbox. Second, hermes serve, the API backend the desktop app and dashboard connect to. In the video below I run Hermes in the cloud for free. Here’s how both kinds of backend work and which to choose.

Hermes Agent terminal backends

Which Hermes Agent backend should you use?

Start with Local if you’re just trying Hermes on your own machine. Switch to Docker as soon as Hermes is running commands you haven’t reviewed — a mistake in a container can’t touch the rest of your system. Use SSH when you want Hermes to work on a VPS or another computer. The cloud sandboxes — Daytona, Modal and Vercel Sandbox — are great for disposable environments and burst work.

The hermes serve API backend

The desktop app and the web dashboard don’t run the agent themselves — they talk to an API backend, hermes serve. On a headless server you can run the backend there and connect a desktop or dashboard to it remotely. It’s separate from the gateway (which runs messaging and cron in the background), but both share the same Hermes home folder.

Running a remote Hermes backend

• Install Hermes on the server with the official script.
• Run the backend and the gateway on the server.
• Reach it securely — Tailscale is the simplest way, instead of opening ports.
• Connect your desktop app or dashboard to it.

That way your agent, memory and jobs live on the server, and your laptop is just a window into it.

Model backends are separate

The terminal backend decides where commands run; the model provider decides which AI does the thinking. You can mix them freely — for example Docker for commands, with Claude on AWS Bedrock or a free Nous Portal model for thinking.

Backend security tips

• Use Docker or a cloud sandbox for untrusted commands.
• Give SSH access to a dedicated user, not root.
• Never expose the API backend to the open internet — use Tailscale.
• Keep approvals on for anything that deletes or sends.
• Update Hermes regularly with hermes update.

Docker backend in practice

With the Docker backend, every command Hermes runs happens inside a container. If it installs a package, deletes a folder or runs a script that goes wrong, the damage stays in the container. That’s why I recommend Docker as soon as Hermes is doing real work with code or files you care about. You can still share specific folders with the container so Hermes can work on your projects, while keeping the rest of your system out of reach.

My Hermes Agent Docker guide walks through the setup.

SSH backend in practice

The SSH backend lets Hermes run commands on another machine — a VPS, a home server or a work computer — while the agent itself runs elsewhere. It’s useful when the files, databases or tools you need live on a server. Use a dedicated user with only the permissions Hermes needs, use key-based authentication, and prefer Tailscale over exposing SSH to the internet.

Choosing a backend by situation

Trying Hermes for the first time: Local. Running scripts or installs you haven’t reviewed: Docker. Managing a server: SSH. Short-lived experiments or bursts of heavy work: a cloud sandbox like Daytona, Modal or Vercel Sandbox. Research computing: Singularity. You can change backends at any time as your setup grows.

Frequently asked: Hermes backends

Can I switch backends later? Yes, at any time, and your memory and skills stay the same. Does Docker slow Hermes down? Slightly for some commands, but the safety is usually worth it. Can I use different backends for different profiles? Yes, which is handy if one bot does risky coding work in Docker while another works locally. Do cloud sandboxes cost money? They’re billed by their providers, so check pricing before running long jobs there. And remember that the backend controls where commands run, not which AI model does the thinking.

The bottom line on Hermes Agent backend

Choose your Hermes Agent backend by risk and location: Local to start, Docker for safety, SSH or a cloud sandbox for remote work, and hermes serve when you want your desktop app to drive Hermes on a server.

If you’re unsure where to start, here’s my simple rule: Local for the first week while you learn, then Docker for anything involving code or installs, and SSH once Hermes needs to work on a server. Revisit the choice whenever you add a new kind of job. Getting the backend right early is one of the easiest ways to avoid nasty surprises as your agent takes on more responsibility, and it costs nothing to change later. Whatever you choose, test it with a harmless command first, like listing a folder, before giving Hermes real work. Then check where the output landed and that nothing outside the intended folder changed. Five minutes of testing tells you exactly how your backend behaves before it matters.

FAQ: hermes agent backend

What is a Hermes Agent backend?

Either the terminal backend where Hermes runs commands, or the hermes serve API backend used by the desktop app and dashboard.

What terminal backends does Hermes support?

Local, Docker, SSH, Daytona, Modal, Vercel Sandbox and Singularity.

What is the safest Hermes backend?

Docker or a cloud sandbox, because commands run in isolation.

Can Hermes run commands on a remote server?

Yes — with the SSH backend.

What is hermes serve?

The API backend that Hermes Desktop and the dashboard connect to.

About Julian Goldie

I’m Julian Goldie — founder of Goldie Agency, best-selling author, and an AI educator with 428K+ YouTube subscribers. I’ve spent 10+ years in SEO and online business and I test every AI agent on my own work before I recommend it. Join the AI Profit Boardroom for the daily builds, or book a free strategy session to talk through yours.

Related reading

• Hermes Agent Background
• Hermes Agent Docker
• Best Hermes Agent VPS

Last updated October 2026. This page is a living guide to hermes agent backend — the facts here move fast and it is updated as they do.