Published: 2026-05-10

Full Setup Guide: Hermes Agent as Your Personal AI Assistant on a Private Server

Chapters / key moments (click to jump — plays here on the page)

Nate Herk's one-hour course covers setting up Hermes Agent from scratch on a private server, demonstrating it as a persistent personal assistant that runs scheduled automations, responds by voice, creates videos, and learns from the tasks you give it. The video includes a direct comparison with Claude Code and OpenClaw, covering when you'd choose each platform and why Hermes fits certain workflows the others don't.

Source video

"Hermes Agent: Zero to Personal AI Assistant (1 Hour Course)" by Nate Herk — Watch on YouTube →

Key Takeaways

Hermes ships with 91 built-in skills out of 684 available — including voice transcription, image analysis, and diagram generation with no additional setup.
Scheduled crons run autonomously: Nate's instance handles daily AI news briefings, YouTube comment monitoring, school community engagement, morning business summaries, and server health checks.
Hermes returns voice responses as audio files alongside text — useful for hands-free updates or client-facing automations.
Hermes self-improves by writing new skills based on tasks you give it — if you ask for a tool that doesn't exist, it researches, installs, and creates the skill itself.
Compares Hermes vs Claude Code vs OpenClaw: Hermes is differentiated by persistent memory, autonomous crons, and self-directed skill acquisition over long time horizons.
Runs on a private server — not a cloud service — giving full data ownership and control over the agent's environment.

What Hermes Looks Like Running in Practice

Nate demonstrates his personal Hermes instance, which runs several ongoing automations: a daily AI news briefing posted to his community, YouTube comment monitoring (with context from his video transcripts, so it can respond on his behalf), morning business summaries, and server health checks. These are crons — scheduled tasks that Hermes runs automatically without prompting. The agent also maintains a persistent model of who Nate is, meaning its responses are personalized across sessions rather than starting fresh each time.

When tasked with creating a video about itself, Hermes autonomously identified the tools needed (including Hyperframes, a tool it didn't have), researched how to install it, asked for permission, installed it, and produced the video — iterating after self-evaluation showed the first version fell short. This pattern — autonomous tool acquisition followed by multi-pass refinement — is the core behavior that makes Hermes different from a standard chat-based assistant.

Hermes vs Claude Code vs OpenClaw

The comparison Nate covers in the course distinguishes between the three platforms on the basis of use case, not quality. Claude Code is optimized for coding tasks, file editing, and development workflows where the agent works within a bounded session. OpenClaw is a persistent local gateway — well-suited for multi-channel communication and extensible via its plugin system. Hermes occupies a different space: it's designed to persist and grow indefinitely on your own server, accumulate skills and memory over months, and run background automations on a schedule without human prompting. The choice depends on whether your primary use case is development, communication routing, or long-running autonomous operation.

What you can actually set up from this

Extracted from the video's own transcript — the specifics the original summary left out.

Configuration shown

Memory files Hermes loads at session start

USER.md   - who you are, your style, preferences, things you dislike
MEMORY.md - environments, projects, business context
SOUL.md   - personality/tone of this Hermes instance (no YAML front matter)
AGENTS.md - project-level context when working in a repo (the CLAUDE.md equivalent)

Template shown in the video. Adapt to your project — do not copy verbatim without reviewing each line.

Reproducible steps

Provision a VPS
The demo used Hostinger KVM 2 (KVM 1 is enough to start, and you can scale up later) with Ubuntu 24.04 LTS. Hostinger also offers a one-click Hermes install.
Choose root install or the one-click Docker image
A direct install lives at the root of the VPS alongside other services. The one-click option runs Hermes in its own Docker container inside the VPS.
Tell it what to remember
Hermes extracts memory automatically, but you should still say things like 'save that to memory' or 'never do this again, put it in USER.md'. Past sessions are stored in a SQL database and searchable with session search.
Install skills
About 91 skills ship built in (the docs listed 684 in total). The Skills Hub has 520+ community skills, 16 of them official Anthropic ones (canvas design, front-end design, skill creator). Paste a hub URL to Hermes and ask it to install.
Create cron jobs in natural language
'Every morning at 6:00 a.m. do X.' Each run starts a fresh isolated session that does not inherit the current chat, runs the skill, and posts the result back to the originating chat.

Gotchas

Cron options mentioned: context_from passes one job's output into another; workdir runs tools from a project folder; a no-agent flag runs just a script with no agent loop.
Cron sessions cannot create more cron jobs, so each cron prompt must be self-contained.
Do not store secrets or temporary task status in memory files.
Skills use YAML front matter for progressive disclosure: only the description loads until the skill is needed.
Contains a VPS affiliate discount code; any VPS provider works.