Engineer-owned racks · Pacific Northwest · AS13725
AN AGENT THAT LEARNS YOU, ON A 99.9% SLA
SELF-HOSTED HERMES AGENT
Hermes Agent is Nous Research's MIT-licensed, self-improving AI agent: it builds persistent memory across sessions, creates and sharpens its own skills, runs scheduled automations, and answers on Telegram, Discord, Slack, WhatsApp, and 20+ platforms. Memory that compounds deserves a home that never powers off. A NoBull Networks VPS is that: RAID NVMe under the memory store, AMD EPYC™ cores capped at 60% load, and a credit-backed 99.9% network SLA.
See plans How deployment works
Illustration of the Hermes Agent learning loop with growing memory and scheduled tasks running on a NoBull Networks VPS

WHY HERMES AGENT WANTS A VPS

A self-improving agent is only as good as the machine underneath it: its memory has to persist, its schedules have to fire, and its blast radius has to stay contained.

Memory Needs A Home

Its knowledge of you is real data
  • Hermes curates memory and skills across every session, so losing the disk means losing the agent you trained
  • RAID NVMe storage keeps recall fast as the memory store grows
  • One free off-node backup on every VPS plan protects what it has learned

Cron Never Sleeps

Scheduled automations, honored
  • Built-in scheduled automations only matter if the box is awake when the clock strikes
  • Credit-backed 99.9% network uptime SLA with redundant switching behind every node
  • Runs as a daemon around the clock while your laptop stays a laptop

A Clean Blast Radius

Full web and terminal access, contained
  • Hermes browses, runs terminal commands, and manages files, which is the point, and the reason it deserves its own box
  • A dedicated VPS is a crisp isolation boundary with nothing else on it
  • Full root for you: give it only the credentials it needs, and keep the kill switch
THE LEARNING LOOP, HOUSED PROPERLY

Chats flow in through the gateway, the agent thinks against your chosen model, and everything it learns stays on your disk.

Personal chats Telegram, WhatsApp, Signal Team chats Discord, Slack, Teams More channels email, SMS, 20+ platforms YOUR NOBULL NETWORKS VPS Hermes Agent learning loop, gateway, cron Memory and skills compounding on RAID NVMe You, privately CLI and dashboard over SSH Frontier models Nous Portal, OpenRouter, OpenAI Any endpoint OpenAI-compatible APIs
Hermes Agent is developed in the open by Nous Research under the MIT license, with official docs and a one-line installer at hermes-agent.nousresearch.com.
RECOMMENDED PLANS FOR HERMES AGENT

The agent itself is lightweight; headroom goes to browsing, skills, and everything you schedule. Every plan is online in under 3 minutes with full root access, and the benchmarks behind it are published.

Performance VPS · P20Solo agent
  • 2 vCPU
  • 4 GB RAM
  • 50 GB RAID NVMe Storage
  • 8 TB Monthly Bandwidth
  • Comfortable for one busy assistant
$16.00/mo
Get this server
Recommended
Performance VPS · P40The daily driver
  • 2 vCPU
  • 8 GB RAM
  • 100 GB RAID NVMe Storage
  • 16 TB Monthly Bandwidth
  • Headroom for browsing and skills
$32.00/mo
Get this server
Performance VPS · P50Heavy automation
  • 3 vCPU
  • 12 GB RAM
  • 150 GB RAID NVMe Storage
  • 24 TB Monthly Bandwidth
  • Room for schedules that never stop
$48.00/mo
Get this server
Model usage is billed by your provider, whether that is Nous Portal, OpenRouter, OpenAI, or any compatible endpoint. Your NoBull Networks invoice stays a flat server price. See the receipts for the benchmarks behind every plan.
An Agent That Compounds

What It Learns Belongs To You

Hermes Agent's pitch is the closed learning loop: it creates skills from experience, improves them while working, and builds a deepening model of you across sessions. That accumulated memory is exactly the kind of data that should live on a disk you control, not in someone else's agent cloud.

This is an unmanaged deployment, said plainly: you install with the official one-line installer and own the agent and its guardrails; our engineers keep the hardware, network, and uptime SLA honest underneath, 24/7.

Treat it like the powerful tool it is: its own server, scoped credentials, and access over SSH.

  • MIT-Licensed Open Source developed by Nous Research
  • Persistent Memory, Protected RAID NVMe plus one free off-node backup
  • Always-On Daemon Home 99.9% network uptime SLA, credit-backed
  • Dedicated Isolation Boundary the agent gets its own box
  • Dedicated IPv4 Included stable connections for every channel
  • Full Root Access your keys, your rules, your kill switch
FROM ORDER TO FIRST LESSON

The official installer does the heavy lifting. Bring a model provider account and one chat app to pair.

Order your VPS

Pick a plan and check out. Your server is online in under 3 minutes with full root access and a dedicated IPv4 address.

Run the installer

SSH in and run the official one-line install script from hermes-agent.nousresearch.com, then follow the onboarding.

Connect model and channels

Add your provider credentials, Nous Portal, OpenRouter, OpenAI, or any compatible endpoint, and pair Telegram, Discord, Slack, or WhatsApp.

Set schedules, stay in charge

Hand it standing jobs with built-in cron, scope its credentials tightly, and keep CLI access to yourself over SSH.

Full installation and configuration documentation lives at hermes-agent.nousresearch.com/docs, maintained by Nous Research.
Frequently Asked Questions
Self-hosting Hermes Agent on a NoBull Networks VPS

Explore the knowledge base or open a ticket in the Console. An engineer answers.

What exactly is Hermes Agent?
Hermes Agent is an open source, MIT-licensed AI agent built by Nous Research. Its signature is the closed learning loop: it curates its own memory, creates skills from experience, improves those skills during use, and carries what it learns across sessions. It also runs scheduled automations and connects to 20+ chat platforms through a unified gateway.
Hermes Agent or OpenClaw?
Both are open source, self-hosted agent gateways, and both run on the same plans here. Hermes Agent leans into the learning loop, persistent memory, and self-improving skills. OpenClaw leans into its channel ecosystem and community plugins. We host both without playing favorites; read both guides and pick the one that fits how you work.
What do I need to get started?
Three things: a VPS (the P20 works, the P40 gives headroom), an account with a model provider such as Nous Portal, OpenRouter, or OpenAI with billing enabled, and optionally a chat app to pair. The official installer is a single command on a fresh Linux server, and the CLI works even before you connect any channel.
What happens to everything it learns?
Its memory and skills live on your server's disk, on RAID NVMe, covered by the free off-node backup included with every VPS plan. That is the practical argument for self-hosting an agent like this: the model of you it builds over months stays yours, portable and deletable on your terms.
Is it safe to run an agent with terminal access?
Respect it like any powerful tool. Give it a dedicated server so whatever it does happens inside its own box, scope its credentials to only what it needs, and keep administrative access behind SSH. A VPS gives you the structural isolation; the operational discipline is yours.
What does it cost to run?
Hermes Agent itself is free under the MIT license. You pay our flat VPS price for the server and your model provider for tokens. Heavier use costs more in model usage, not hosting; your NoBull Networks invoice does not move.
READY TO GROW YOUR OWN AGENT?

Order a VPS, run the one-line installer, and start the learning loop today.

Performance VPS
P40 at $32.00/mo Get this server
All VPS Plans
Compare every tier See VPS plans
More Solutions
See what else fits Browse solutions