Hermes Agent Server Requirements: The Missing Numbers
The docs have no system requirements page (issue 73754). From measured reports and our production fleet: 1 GB is a demo, 2 GB is the always-on floor, 4 GB is comfortable, local models start at 8 GB, 2 cores suffice, and 20 GB of disk keeps it boring.
The Hermes Agent documentation answers almost everything except the question people ask first: what machine does this need? Issue 73754, open since July 29, 2026, states it plainly: the installation page's prerequisites cover software only, and a grep across the getting-started docs finds no RAM, disk or CPU guidance at all. This page assembles the missing numbers from the measurements that exist: figures reported in the project's own issue tracker, the official platform list, and a fleet of Hermes servers we operate in production.
The two measurements that set the floor
Two numbers from the issue tracker do most of the work. First, the gateway, the always-on process, was measured growing from roughly 300 MB at a fresh start to between 1.4 and 1.6 GB over about 20 hours of light use, in issue 58817, July 5, 2026. Second, the dashboard's web build alone peaked at 387.9 MB and got the process killed on a 960 MB server, logged as an out-of-memory kill in issue 14898, April 24, 2026.
What this means for you: a 1 GB machine sits under the waterline of the gateway's own measured growth. It can boot Hermes and it lives one busy afternoon away from the out-of-memory killer. Treat 1 GB as a demo, and 2 GB as the practical floor for an agent that stays up.
The sizing table the docs skip
| Setup | RAM | Why |
|---|---|---|
| Demo, short sessions | 1 GB | Boots, then flirts with the OOM killer issue 14898 documented |
| Always-on agent, cloud models | 2 GB | Covers the gateway's measured 1.4 to 1.6 GB growth with headroom |
| Agent plus dashboard plus room to grow | 4 GB | Build spikes, browser automation and updates stop competing |
| Local models on the same box | 8 GB and up | The docs' own local model guide starts at 8 GB for 3B-parameter models, cited in issue 73754 |
CPU is the quiet part of the story: on our own production fleet, Hermes agents run around the clock on 2-core dedicated servers without strain, measured through July and August 2026. Disk deserves more respect than it gets. The runtime installs Python 3.11 and Node.js v26 per the official installation page, browser automation adds a full Chrome build, and issue 58172 measured update leftovers alone reaching 884 MB after months of hermes update. Twenty gigabytes of disk keeps all of that boring.
Operating systems, from the official list
The official platform page lists tier-one support for macOS on Apple Silicon, Windows 10 and 11, Linux and WSL2 on x86_64 and aarch64, and Docker containers, with Android via Termux and Nix as best-effort tiers. For a server, that makes an ordinary Ubuntu LTS on x86_64 the well-trodden path, and our troubleshooting section covers the Ubuntu install traps separately.
Size it in one minute
- Decide where models run. Cloud models keep the box light; local models move you to the 8 GB league before anything else matters.
- Pick RAM off the table above. 2 GB always-on floor, 4 GB comfortable.
- Give it 20 GB of disk so the runtime, the browser and months of updates never fight.
- Stay on the tier-one list: Ubuntu LTS on x86_64 is the path the reports and fixes assume.
Or skip the arithmetic: our managed Hermes Agent hosting starts on a dedicated EU Starter server with 2 cores, 4 GB of RAM and 15 GB of disk, and the 7-day free trial needs no card. For the full 20 GB recommended above, the Hobbyist plan has 25 GB. For what the machine costs month to month, the Hermes Agent cost guide breaks it down.
Sources: Hermes Agent issues 73754, 58817, 14898 and 58172, the official installation and platform support pages, all read on August 29, 2026, and measurements on our own production fleet from July and August 2026.
Frequently asked questions
- What are the minimum system requirements for Hermes Agent?
- 2 GB of RAM, 2 CPU cores and 20 GB of disk for an agent that stays on and uses cloud models. 1 GB boots, but issue 14898 recorded an out-of-memory kill during the dashboard build on a 960 MB server, and issue 58817 measured the gateway alone growing to 1.4 to 1.6 GB over 20 hours. 4 GB gives headroom for the dashboard and browser automation.
- How much RAM does Hermes Agent need?
- The gateway alone was measured growing to 1.4 to 1.6 GB over 20 hours in issue 58817, so 2 GB is the practical floor for an always-on agent on cloud models. 4 GB adds comfortable headroom for the dashboard and browser automation.
- Does Hermes Agent run on a 1 GB VPS?
- It boots, and issue 14898 documents the risk: the dashboard build alone peaked at 387.9 MB and drew an out-of-memory kill on a 960 MB server. Treat 1 GB as a demo tier.
- Which operating systems does Hermes Agent support?
- Tier one per the official platform page: macOS on Apple Silicon, Windows 10 and 11, Linux and WSL2 on x86_64 and aarch64, and Docker. Android via Termux and Nix are best-effort tiers.
- How much disk space does Hermes Agent use?
- The runtime installs Python 3.11 and Node.js v26, browser automation adds a full Chrome build, and issue 58172 measured update leftovers alone at 884 MB after months. 20 GB keeps everything comfortable.
Related guides
- OpenClaw dreaming: what it does and how to switch it offOpenClaw dreaming tidies your agent's memory every night at 3 in the morning. A beginner guide with a screenshot for every step: what it does, what it costs and where the switch is.
- ChatGPT Sign-In on Your Agent Waits or Times Out: Turn On Device Code Sign-InA hosted agent signs in to ChatGPT with a one-time code, and ChatGPT accepts those codes only after one switch in its security settings is on. This guide shows the switch, how to sign in again on the crab (OpenClaw) and on Hermes, and what each message means.
- Your Crab or Hermes Stays Silent After Setup: Connect an AI Model FirstA freshly built server runs the agent, and the agent waits for an AI model. Until a model answers one test request, every chat stays quiet. This guide shows where that one step lives for the crab (OpenClaw) and for Hermes, where an API key comes from, and what to check when the silence stays.
- Best AI Models for OpenClaw and Hermes Agent in 2026OpenClaw and Hermes Agent both need an AI model to think with. This guide lists twelve ways to connect one, by API key, by subscription sign-in or on your own machine with Ollama, with what each costs in a month.