agentos.guide › blog

Hermes Agent Backend: The Quick-Start, Every Option, And Where Each One Fits In An Agent OS

By Julian Goldie · 7 October 2026 · agentos.guide

A Hermes Agent backend is the layer that does the work behind the chat, and the main one is the terminal backend, which decides where the agent's shell commands run.

The official docs list seven terminal backends, which are local, Docker, SSH, Modal, Daytona, Vercel Sandbox and Singularity.

The same word also covers the model provider, the web search provider, the browser provider, the memory provider, and Hermes itself running as a backend for other apps.

Everything here comes from the official Hermes docs and the read-only help output on my own install, which is version 0.21.4.

Here is the full map first, then a seven-step quick-start and where each backend sits in an Agent OS.

What a Hermes Agent backend is

Hermes Agent is the free, open-source AI agent from Nous Research, and it is released under the MIT licence.

The docs use the word "backend" for six different layers, which is why the term confuses people.

Here is every layer in one table, with the options the docs list and the command that sets each one.

The terminal backend is the one most people mean, so the quick-start begins there.

Hermes Agent backend quick-start

These seven steps take a new install from the default to a sandboxed setup.

The terminal setting is stored under terminal.backend in ./.hermes/config.yaml, so you can also edit the file by hand.

If the new backend fails, you set the value back to local and test again.

My own config is set to local, which is the default.

The seven terminal backends

The docs say each terminal backend determines where the agent's shell commands actually execute.

The local backend uses your real home folder, so tools like git and npm find the logins they already have.

The Docker backend starts one long-lived container and sends every terminal, file and code call into it.

A package the agent installs once is still there on the next command and after a restart.

You get a fresh container for every chat when you set container_persistent: false.

You cut the container off from the internet when you set docker_network: false.

The SSH backend keeps one shell alive on the remote host, so the working directory survives between commands.

The three cloud sandboxes save your files between sessions, and the docs say they don't save running processes.

If you want a server for the SSH route, I've ranked hosts in the best Hermes Agent VPS.

🔥 Want the exact Hermes backends in the Agent OS setup I run inside my Agent OS?

Inside the AI Profit Boardroom you get the installable Agent OS, step-by-step video tutorials, four coaching calls a week and 3,400+ members building real automations.

→ Get access here

The model backend

The second meaning of a Hermes Agent backend is the provider that does the thinking.

You add a provider with hermes model in your terminal, and you switch between providers you've already added with /model inside a chat.

The docs call Nous Portal the recommended way to run Hermes, because one login covers 300+ models and the Tool Gateway.

The docs also warn that Ollama starts with a small context window, so you set context_length in the config for a local model.

You can add a backup chain with hermes fallback, which moves to another provider when the main model fails.

The provider list in the docs is far longer than my table, and it includes Google Gemini, DeepSeek, xAI, Vertex AI and Azure AI Foundry.

Bedrock has enough moving parts that it gets its own guide, Hermes Agent on Bedrock.

Web, browser and memory backends

Hermes has two web tools, which are web_search and web_extract.

Firecrawl is the default web backend, and the docs also list SearXNG, Brave Search, DuckDuckGo, Exa, Parallel, Tavily, Perplexity, Keenable, xAI and OpenAI Native.

A fresh install works with no key, because requests rotate across the free tiers of Exa, Parallel, Firecrawl and Keenable.

You can split the job with web.search_backend and web.extract_backend, so a free search provider sits next to a different extract provider.

The browser toolset has its own backends, which are Browser Use, Browserbase, Firecrawl, Camofox, Lightpanda, your own Chrome over CDP and a local Chromium.

You choose one in hermes tools under Browser Automation, and the choice is stored as browser.cloud_provider.

The full toolset is covered in the Hermes Agent browser guide.

Memory has a built-in layer in MEMORY.md and USER.md, and that layer is always on.

You can add one external memory provider with hermes memory setup, and only one can be active at a time.

My ranked list of those providers is in the best Hermes Agent memory.

Using Hermes Agent as a backend

The last meaning flips the direction, because Hermes becomes the backend for something else.

The API server listens on http://127.0.0.1:8642 and speaks the OpenAI Chat Completions and Responses formats.

The docs name Open WebUI, LobeChat, LibreChat, NextChat and ChatBox as frontends that can connect to it.

The agent answers with its full toolset, so a chat app gets terminal, files, web search, memory and skills behind it.

The docs say hermes mcp serve is stdio-only today.

The hermes serve command binds to 127.0.0.1 by default, and a public bind always requires an auth provider.

The docs draw a clear line between two of these: you use the API server when you want the agent, and you use the proxy when you only want the model.

🔥 Want the exact Hermes backends in the Agent OS setup I run inside my Agent OS?

Inside the AI Profit Boardroom you get the installable Agent OS, step-by-step video tutorials, four coaching calls a week and 3,400+ members building real automations.

→ Get access here

Where each backend fits in an Agent OS

An Agent OS is one dashboard over the same agent, so every room in it leans on one of these backends.

Hermes profiles make this easy, because each profile is its own isolated Hermes with its own config.

That means one profile can run a local Ollama model while another runs a cloud model.

My default profile points at a local model, and I keep separate profiles for other models.

Herald OS shows the same idea in a settings page, where you choose where agents run tasks, including Docker.

Security trade-offs: local versus Docker

This is the choice that matters most.

On local, the docs say the agent has the same filesystem access as your user account.

Hermes offsets that by checking commands against a list of dangerous patterns and asking you before it runs a match.

On Docker, the container starts with all Linux capabilities dropped, no privilege escalation and a limit of 256 processes.

The docs recommend docker, modal, daytona or vercel_sandbox for production gateway deployments.

A prompt only protects you when a human is there to read it, and a container protects you when nobody is watching.

There is one catch with Docker, which is that any secret you list in docker_forward_env can be read by code inside the container.

Performance and cost

• The local backend has no sandbox to start, so nothing sits between the agent and your shell.
• The Docker backend reuses one running container instead of building a new one for each command.
• The SSH backend reuses one connection and one persistent shell.
• Container backends default to 1 CPU core and 5 GB of memory, and you can raise both in the config.
• Every command has a 180 second timeout by default.

The docs also warn that slow shell startup files add delay to every command, because the agent starts many shells.

Hermes is free, and local, Docker, SSH and Singularity cost nothing beyond your own hardware or server.

Modal, Daytona and Vercel Sandbox are paid cloud services with their own pricing, so check each one before you commit.

Your model provider bills separately, unless you run a local model.

Once the backend is settled, Hermes Agent background tasks shows how to keep the agent working after you close the window.

Common mistakes

• People switch to Docker while Docker isn't running, and the terminal tool then fails.
• People set the SSH backend with a host and no user, and both values are required.
• People expect a cloud sandbox to keep a dev server alive, and only the files are saved.
• People try to add a new provider with /model, and that job belongs to hermes model.
• People forward every secret into the container, which removes the point of the sandbox.

FAQ

What is a Hermes Agent backend?

It most often means the terminal backend, which decides where the agent's shell commands run. The docs list seven: local, docker, ssh, modal, daytona, vercel_sandbox and singularity. The word also covers the model provider, web provider, browser provider and memory provider.

How do I change the Hermes Agent backend?

Run hermes config set terminal.backend followed by a name such as docker. The value is saved under terminal.backend in ~/.hermes/config.yaml. Run hermes model for the provider, hermes tools for web and browser, and hermes memory setup for memory.

Is Docker safer than the local backend?

Yes. On local the agent has the same file access as your user account. On Docker, commands run in a container with all capabilities dropped, no privilege escalation and a 256 process limit.

Can I run Hermes Agent with a local model backend?

Yes. Choose Custom endpoint in hermes model and enter http://localhost:11434/v1 for Ollama, or pick LM Studio. The docs advise setting context_length, because Ollama starts with a small context window.

Can Hermes Agent be the backend for another app?

Yes. The API server exposes Hermes as an OpenAI-compatible endpoint on port 8642 for frontends such as Open WebUI. Hermes also runs as an ACP server with hermes acp and as an MCP server with hermes mcp serve.

Do I need an Agent OS to use these backends?

No. Every backend here is set from the Hermes terminal. An Agent OS puts the same agent behind one dashboard, so the backend you choose applies to every room in it.

Final word on the Hermes Agent backend

Start on local, switch to Docker the day another person can message your agent, and give each profile the model it needs.

Do those three things and you'll have the right Hermes Agent backend under your whole Agent OS.

Related guides on agentos.guide

→ The Best Hermes Agent Setup

→ The Best Hermes Agent Tools

→ Herald OS: The Free Hermes Agent OS

→ Build Your Own AI Agent OS

📺 Video notes + links to the tools 👉

🎥 Learn how I make these videos 👉

🆓 Get a FREE AI Course + Community + 1,000 AI Agents 👉

← all guides on agentos.guide