All Agents
Hermes Agent

Managed Hermes Agent Hosting

Personal AssistantMIT230k stars45k forks

Managed Hermes Agent hosting runs Hermes on infrastructure we operate, instead of on your laptop or a VPS you maintain. Mantlecore provisions the virtual machine, keeps the host running, issues HTTPS and takes a daily off-site backup. You choose the model, connect the channels, and decide what the agent is allowed to reach.

Plans start at $19 per month and include a 7-day free trial. A payment card is required at signup, and the trial itself is not charged. New accounts also receive $2 of managed AI credit.

Requirements

Min CPU2 vCPU
Min RAM4 GB

Starting at $19/mo with VM isolation, persistent storage, and daily backups.

About Hermes Agent

Hermes is an open-source personal AI agent by Nous Research, designed to be your agent across every platform you already use. Bridges 20 messaging platforms (Slack, Telegram, WhatsApp, Signal, Email, Discord, Teams, Google Chat, and more) through a single gateway. Persistent identity via SOUL.md, MEMORY.md, and USER.md — your agent remembers conversations across sessions and grows a procedural-memory library of self-authored skills. Built-in curator periodically refines its own skills library. Multi-provider LLM support routed through our LiteLLM proxy — pick any model, swap anytime. Durable multi-agent kanban and persistent goal loops keep long-running work moving across sessions.

What You Get With Hermes Agent

Features

20 messaging platforms — Slack, Telegram, WhatsApp, Signal, Email, Discord, Teams, more
Personal AI assistant across Slack, Telegram, WhatsApp at once
Persistent identity — SOUL.md, MEMORY.md, USER.md that survive across sessions
Build a long-running agent that learns your team's preferences
Self-authored skills — agent writes its own procedures after complex tasks
Replace a virtual assistant — your Hermes pairs with your email and calendar
Built-in curator refines the skills library on a weekly cadence
Autonomous research agent with persistent memory of prior work
Multi-agent Kanban with heartbeat / zombie detection
Multi-agent orchestration via the built-in Kanban for complex projects
Embedded TUI in dashboard — the canonical Hermes experience, in your browser
OpenAI-compatible API on a separate port for integrating with your IDE, scripts, or curl
Provider-agnostic — Anthropic, OpenAI, Google, Nous Portal, OpenRouter, all routed through LiteLLM

What we manage, and what stays yours

Managed hosting is a split of responsibility, so it is worth writing the split down rather than leaving it implied.

Deployment

Mantlecore handles
Provisioning the virtual machine and the Hermes environment on it
You control
Which agent and which plan you deploy

Isolation

Mantlecore handles
One virtual machine per agent, each with its own guest kernel
You control
Tool permissions, credentials, and which external services the agent can reach

HTTPS

Mantlecore handles
Certificates and TLS for your agent's endpoint
You control
Which channels and integrations you connect

Storage

Mantlecore handles
Persistent VM storage that survives stop, start and restart
You control
The files and agent data created inside the workspace

Backups

Mantlecore handles
One off-site backup per day, retained for 7 days on Starter and 30 days on larger plans
You control
When to restore, and any recovery needed outside the VM

Snapshots

Mantlecore handles
An automatic snapshot taken immediately before any version change
You control
Your own manual snapshots — 1 on Starter, 3 and 5 on the larger plans

Updates

Mantlecore handles
Publishing verified versions, snapshotting before the change, and rolling back automatically if the post-update health check fails
You control
When to apply an update, and whether to roll back afterwards

Models

Mantlecore handles
Managed model access, or support for your own provider API key
You control
Provider, model and budget

Why the VM boundary matters

Hermes can run terminal commands, work with files, drive a browser and reach external services. That makes the runtime boundary part of the hosting decision, not an implementation detail.

A conventional container shares the host kernel with everything else on the machine. Each agent here gets its own virtual machine and its own guest kernel instead, so agent workloads do not share a kernel with one another.

That reduces one class of risk. It does not make every agent action safe. You still want scoped credentials, restricted tool permissions, and human approval for anything irreversible — a machine restore cannot unsend a message, reverse a payment, or undo a write that already reached someone else's database.

Your Hermes state persists

Hermes accumulates configuration, sessions, memory, and a library of skills it writes for itself. Persistent storage keeps that available across restarts rather than resetting the agent to a clean image each time.

Machine state is broader than agent memory, though. Installed packages, service configuration, permissions and scheduled processes all affect whether an agent comes back genuinely usable. Backups here capture the machine, not just a memory file.

Choose managed hosting when the agent is the work

Self-host Hermes if you already operate Linux comfortably, want full root control, or need to customise parts of the stack a managed service does not expose. That is a reasonable choice and this page is not going to argue you out of it.

Use Mantlecore when you want Hermes running on a persistent, isolated VM and would rather not provision a server, configure TLS, or build and test a backup path yourself.

Managed hosting does not make Hermes more capable. It removes the infrastructure work between choosing an agent and keeping it running.

Questions people actually ask

What is managed Hermes Agent hosting?
It means Hermes runs on infrastructure operated by a provider rather than on your own laptop or VPS. A provider should tell you exactly which parts it manages — provisioning, the host, HTTPS, storage, backups and the update path — and which parts stay yours. Here, you keep control of the model, the connected services and the agent's permissions.
Does each Hermes Agent get its own VM?
Yes. Every agent deployment gets its own virtual machine with its own guest kernel, rather than a container sharing the host kernel with other workloads. That boundary does not replace scoped credentials, network policy, or approval for high-impact actions.
Will my Hermes memory and skills survive a restart?
Yes. The workspace is on persistent storage, so memory, skills and configuration are not discarded on a normal restart. A daily off-site backup is a separate recovery control, retained for 7 days on Starter and 30 days on the larger plans.
Can I use my own model API key?
Yes. You can bring a supported provider key or use managed model access and pay from credit. Either way, your prompts and the relevant context go to whichever model provider you select — that traffic leaves the VM by design.
Can a backup reverse anything Hermes has done?
No, and this is worth being precise about. Restoring a backup recovers machine state as captured at that point. It cannot reverse an action that already completed somewhere else — a sent message, a payment, or a write to an external database stays done.
Who applies updates, and what happens if one breaks?
You decide when to update; nothing is upgraded underneath you. When you apply one, a snapshot is taken automatically first, and if the health check afterwards fails the update rolls back on its own. You can also roll back yourself later.
How much does managed Hermes hosting cost?
Plans start at $19 per month, with a 7-day free trial and $2 of managed AI credit on new accounts. A payment card is required at signup, though the trial is not charged. Check the pricing page for current terms.
Should I use Mantlecore or a self-managed VPS?
Use a VPS if you want root access and are happy owning patching, networking, process supervision, monitoring, backups and restore testing. Use Mantlecore if you want a persistent, isolated Hermes VM without operating the host around it.

Keep reading

Ready to deploy Hermes Agent?

Its own VM. Persistent workspace. Managed backups and update recovery.