Not always. If you work alone and want to get to know Hermes first, Hermes Desktop can run locally on your existing computer. A separate Linux computer becomes useful when several people use the same agent. It also helps when tasks need to continue after laptops close, or when you want to manage files, tools, and integrations centrally.
Linux is not the goal in itself. You are mainly choosing where Hermes works each day: on one employee's computer or on a separate, managed computer for the team. Below, we compare both options and explain which hardware really makes a difference.
For one person, the simplest option is often Hermes Desktop on their own work computer. The official Windows guide is clear about it: "Hermes runs natively on Windows 10 and Windows 11 — no WSL, no Cygwin, no Docker." So you do not need an extra Linux layer on Windows. You can also start locally on macOS and Linux. You get access to your own projects and tools without buying a separate computer first.
This approach fits when you mainly work with the agent yourself and your computer is usually on. You do not need a shared business environment yet. You can gain experience with real tasks and invest only once you know which workflows provide value.
A managed Linux computer becomes useful when the agent can no longer be tied to a single laptop or user. The team then works together through a channel such as Discord, Slack, or Buzz. Hermes runs on that Linux computer, the node, and carries out the tasks there. Hermes Desktop can also connect separately to the same node through a Hermes gateway. Hermes Desktop remains the local starting point for one person; teams choose the channel that suits them.
Your laptop is no longer where every task has to run. The Linux computer can open browsers, run scripts, process documents, and maintain projects. Meanwhile, employees keep using their own computers. When someone closes a laptop, the central work environment does not have to stop too.
Context also becomes easier to administer. Project files, work instructions, learned procedures, and selected integrations live in one controlled environment. The business decides where it stores the authoritative version of source code and documents, such as GitHub or an internal Git server. The Linux computer can store local working copies and temporary files for task execution.
For a small team, this mainly reduces administrative burden. Everyone uses the same configured environment. You no longer need to arrange updates, monitoring, and recovery separately on every work computer.
The table shows, item by item, where Hermes Desktop on your own computer and a managed Linux computer differ.
| Aspect | Hermes Desktop on your computer | Managed Linux computer |
|---|---|---|
Best starting point | One person starting locally | A team or recurring work |
Interaction | Hermes Desktop on the same computer | Discord, Slack, or Buzz as a shared work environment |
Availability | Depends on the work computer and local installation | Can remain available independently of employees' laptops |
Projects and context | Configured per work computer | One centrally administered project environment |
Local workload | Browsers, builds, and scripts use the work computer | The Linux computer handles task execution |
Administration | The user or internal IT administers the installation | Voltti handles installation, monitoring, updates, and recovery |
A rented virtual private server (VPS) at a hosting provider can also be a permanent workplace for an AI agent. The agent, your files, and your integrations then live in that provider's data center, not on your own network. Such a server is also reachable from the internet by default. Voltti never installs Hermes on publicly reachable servers. An installation always sits behind a firewall, and you can only reach it through the business network or a VPN such as Tailscale. If you want files, tools, and business knowledge to stay on your own network, your own Linux computer is a better fit. You can read the broader trade-off in local AI vs. cloud AI.
Linux works well as a permanent work environment for software that carries out tasks independently. Automation, browsers, development tools, and background processes are well supported. Permissions and files can also be administered centrally, without an employee's everyday computer doubling as a server.
Hermes works well on Windows for local use. Linux becomes attractive when you want a neutral, stable environment separate from the team's different Windows, macOS, or Linux computers.
For workstations that an administrator also uses directly, Voltti often chooses Omarchy. This Linux environment combines a polished desktop with a strong foundation for software development and automation. That makes the computer practical to maintain both remotely and on site.
An AI agent does not need a striking desktop to work well. For us, Omarchy's value is its consistent work environment and the ability to intervene quickly when needed. A simpler Linux installation may be more suitable for a server without a desktop. That is why we choose the Linux environment for each setup.
On a managed Linux computer, Hermes, files, tools, business knowledge, and integrations can run locally. That does not mean every language model also runs on that computer. When using an external model provider, a model request may include your prompt, conversation context, parts of your files, attachments, and tool results. During the pilot, we also agree where chat messages, connected services, and backups are stored. To see which parts stay local and which go to a provider, read what stays local in a hybrid AI setup.
Small local models may run on a CPU. They are slower, though, and do not match strong models for productive team use. Strong, practically useful local models usually require separate GPU hardware. That is a different decision with its own costs, performance, and administration. On our page about local AI for businesses, we explain the difference between a local agent and a local language model in more detail.
As long as an external provider supplies the language model, a Linux computer for Hermes does not need to be an expensive AI server. Most local load comes from browser automation, software builds, file processing, and tools running at the same time. A practical base for a team usually includes:
a modern processor with enough capacity for several tasks at once;
32 GB of RAM, with an upgrade path to 64 GB for heavier use;
at least 1 TB of fast SSD storage (NVMe) for projects and temporary working files;
a reliable wired network connection and cooling for sustained use;
replaceable components and room for additional storage or memory.
A lightweight pilot for one user can start smaller. For comparison, our Raspberry Pi build with Hermes and Home Assistant runs as a home agent with just 8 GB of RAM. That works because the language model runs externally. The capacity you need depends mainly on what runs at the same time: browser automation, software builds, and other tools. The number of employees matters less. Strong local models or heavy local image and video tasks call for a separate GPU solution. If you also want to run a language model locally, read what computer you need to run AI locally.
Voltti installs Hermes on the hardware you already have or choose; a computer supplied by Voltti is one option. We configure Hermes around the selected business processes, connect only the necessary systems, and build up work instructions and business knowledge. Updates, monitoring, and assistance then form the ongoing support service.
Encrypted backups go to a separate location in the customer's network, never to the Linux computer itself. They include Hermes settings and sessions, business knowledge, and project data that you cannot easily recreate or retrieve from Git. Local changes not yet pushed to Git are included too. We agree retention periods with the business.
Voltti monitors whether backups succeed and tests recovery periodically. A backup file alone is not a proven recovery path. To see how such a recovery path works, read what happens to your AI environment if hardware fails.
A dedicated computer makes administration easier, but it does not automatically make the environment safe. Only approved users get access. Integrations receive the permissions needed for the agreed task. We configure mandatory confirmation for important or irreversible actions. Before you connect business data, also work through our AI safety checklist.
Together with the business, we define which information stays local and which context an external model provider may process. We also agree where human oversight is required. Read more about our principles for security and privacy.
The move suits a business that wants a small team to use the same agent. Recurring tasks, or a wish to manage projects and integrations centrally, are good reasons too. It also makes sense when the agent needs to remain available without depending on one employee's laptop.
If you currently work alone and use Hermes occasionally, Hermes Desktop on your existing computer is usually the most sensible starting point. If usage grows to a team, we move the existing knowledge and configuration to a managed Linux computer. The team then works with it through a channel such as Discord, Slack, or Buzz. Strict separation between departments, a fully offline language model, or heavy local models requires a more extensive architecture than a standard workstation.
Voltti builds these pilots for businesses in Belgium and the Netherlands. The page about the Hermes AI agent explains what Hermes can do. Our pilot approach shows how we first test one specific workflow under controlled conditions.
No. Hermes runs natively on Windows 10 and 11 and can also be used on macOS or Linux. For one person, Hermes Desktop on an existing computer is often the simplest start. Linux is mainly useful as a stable, central work environment for a team or recurring tasks. On Windows, you do not need WSL or Docker for it.
Voltti handles the agreed installation, monitoring, updates, and recovery process. Encrypted backups go to a separate location in the customer's network. We agree retention periods in advance, monitor backup success, and test recovery periodically. A backup is never stored on the Linux computer itself. You decide how often backups run when we set up the installation.
The team works in shared channels through a service that suits them, such as Discord, Slack, or Buzz. Hermes runs on the managed Linux computer and carries out tasks there. Employees get a straightforward shared work environment, while the central computer is administered separately. Hermes Desktop can also connect separately to the same node through a Hermes gateway.
Not automatically. Files, tools, business knowledge, and the agent can run locally. An external model request, however, may include your prompt, conversation context, parts of your files, attachments, and tool results. Working copies and temporary files may also be stored on the Linux computer for execution. During the pilot, we define where source code, documents, chat messages, connected services, and backups are processed and stored.
Not when Hermes uses an external model provider. Small local models may run on a CPU, but are generally slower. Strong local models for team use or heavy local image and video tasks usually require a separate GPU. For the agent itself, a modern processor, 32 GB of RAM, and fast SSD storage are usually enough. If you do want to run a language model locally, that model determines which computer you need.
