AI Hosting Hermes-ready AI hosting

Hermes hosting, for an agent that works while you're not watching.

Hermes is a family of open-weight models from Nous Research: Hermes 3 is current, Hermes 4 is rolling out. Hermes Agent is the runtime built on top of it: it remembers past sessions, schedules tasks for later, delegates work to subagents, and uses real tools to browse, work with files, and run code. It runs on our AI Hosting plans today, staying online and under your control 24/7.

View AI plans Plan a Hermes deployment

Hermes

Private assistant layer

Ollama

Model runtime ready

Open WebUI

Chat interface ready

NVMe

Fast AI storage

AI control plane Live on AI Hosting

Hermes

assistant layer

Ollama

model runtime

Open WebUI

chat interface

Workers

automation

Compute

Memory

NVMe

$ ai-hosting route hermes

$ runtime prepare --ollama --assistant

$ interface attach --open-webui

$ worker scale --queue=chat

Why self-host it

An agent that stays on the clock after you log off.

A VPS gives Hermes Agent a place to live where it stays online, stays under your control, and stays cheap enough to leave running 24/7. Pick a plan sized to your team and how many subagents you want running at once, then upgrade in place. The Agent keeps its memory and schedules across the move.

View AI plans Talk through a Hermes build

Keeps working while you're offline

Runs scheduled jobs overnight and keeps memory of past sessions on local disk. It doesn't reset every time you close the tab.

Flat bill, not per-token

A heavy week running the Agent costs the same as a quiet one. The model runs on your server, so usage doesn't grow the bill.

Subagents and delegation

Hermes Agent doesn't just chat. It delegates work to subagents and uses real tools to browse, work with files, and run code.

NVMe-backed data path

Keep conversation history, prompts, logs, and app data on fast storage with an upgrade path as usage grows.

What it's for

What people build with Hermes.

Almost anything, but here are four shapes the build usually takes once a team actually puts the Agent to work.

Coding agents

Pair the Agent with a coding model for review, refactors, and small features against a private repo.

Internal apps

Drop Hermes behind an internal app so the team can use it without prompts leaving your network.

Research and knowledge work

Pull notes, browse sources, and summarize long inputs. Yesterday's research informs today's session.

Always-on automations

Cron jobs that think: schedule weekly digests or end-of-day reports that run on their own.

Optional AI stack

Pair Hermes with the tools an assistant needs.

A private assistant leans on a chat interface, a model runtime, and workers for anything beyond conversation. AI Hosting can support all of it under one plan.

Open WebUI Private AI chat

Ollama Model runtime

AnythingLLM Team AI workspace

LocalAI OpenAI-compatible API

Dify AI app builder

Flowise Visual AI flows

n8n AI Workflow automation

OpenClaw Agent automation

Langflow Agent pipelines

vLLM Inference server

Qdrant Vector database

Weaviate AI-native search

Milvus Large vector search

ComfyUI Image workflows

Whisper Speech to text

JupyterLab AI notebooks

What's included

The Hermes model is the brain. This is everything around it.

Interface, model runtime, memory, and automation, all on one plan instead of scattered across separate services. Swap the model without rebuilding the Agent.

Interface

Host the Hermes Web UI over an SSH tunnel, or a custom chat frontend for team assistants and internal tools.

Model runtime

Run Ollama behind the Agent, or point it at a hosted model API instead. Swap models without rebuilding the setup.

Memory

Session history and context live on your own disk, so the Agent picks up where the last session left off.

Subagents and scheduling

1 Gbit/s port speed

3 snapshots included

DDoS protection included

Firewall included

IPv4 + IPv6 included

Getting started

From plan to a running Hermes Agent.

Install it on a fresh plan, or add it to a server you already have.

Pick a plan sized for your team

One person running the Agent daily fits comfortably on a smaller plan. Multiple subagents or heavier pipelines need more RAM and cores.

Install the Agent and connect a model

SSH in, point the Agent at a model, and connect the integrations you want. Check the setup with the Hermes Web UI over an SSH tunnel.

Start building

Give it scheduled jobs, subagents, and tool access, then leave it running.

Sizing it right

One person to a full team, on the same plan family.

Start with one person running Hermes Agent for daily work, then upgrade in place as teammates and subagents join. The Agent keeps its memory and schedules across the move.

Personal or power use

One person running the Agent daily across a handful of integrations.

Team deployment

Multiple subagents running at once, or complex agent pipelines under heavier parallel load.

GPU planning

Running a larger model locally?

Hermes Agent itself runs comfortably on standard plans. No GPU needed. Self-hosting a larger underlying model for faster or heavier workloads needs dedicated GPU infrastructure. Contact us so we can scope it.

Discuss GPU hosting

AI hosting FAQ

Questions before you build on Hermes?

These answers explain how Hermes Agent fits with other AI tools, model choice, GPU needs, and support.

What's the difference between Hermes AI and Hermes Agent? The model versus the runtime built on top of it.

The Hermes model is the brain. Hermes Agent is everything around it: memory across sessions, scheduled tasks, subagent delegation, and the tool layer that lets it browse, read files, and run code. You can swap the model without rebuilding the Agent.

Which model does the assistant run on? Your choice: a self-hosted runtime or a hosted API.

You can run a self-hosted model runtime such as Ollama, or connect the assistant interface to a hosted model API instead. Both approaches work on the same hosting plans; the right choice depends on privacy needs, cost, and how much control you want over the model.

Can I integrate Hermes with GitHub or Tailscale? Yes to both.

Clone the Hermes GitHub repos onto your server, pull updates as new versions land, and use upstream recipes or your own forks. Tailscale puts the Agent on a private mesh so only your own devices can reach it, useful if you don't want it exposed on the open internet.

Can I run Hermes Agent on a VPS? Yes. It's built for exactly that.

Hermes Agent is built to run on a normal Linux server. Pick a plan that fits your workload and how many subagents you want running at once, then install it and connect it to a model.

Does this include GPU hosting? Not by default. Larger self-hosted models need a GPU-backed plan.

The plans on this page cover the interface, runtime, memory, and worker layers around a lighter assistant model. If you plan to self-host a larger model for heavier workloads, contact us so we can scope a GPU-backed option.

Can I run this under my own domain? Yes. That is the preferred production path.

You can connect your own domain or subdomain, add SSL, and run the assistant interface behind your own access rules. This is better for trust, team access, and internal tool adoption.

Start building

Build your private assistant on hosting ready for AI work today.

Choose an AI Hosting plan, connect your assistant interface and model runtime, and add automation workers as the project grows.

Compare AI plans Talk to sales