Digital Product · Instant Download

Build Your Own
Autonomous AI Agent OS

The complete, battle-tested playbook for standing up a persistent AI agent stack — Hermes, Paperclip, VPS — that runs 24/7, talks to you on Telegram, and works while you sleep. For $5/month in infrastructure.

Get the Blueprint
$97 One-time · No subscription
40+ hours of research distilled Works on any VPS — Hetzner CX22 recommended Weekend setup, lifetime payoff
sovereign-stack — hermes gateway
$ hermes gateway start --channels telegram
# Hermes gateway active — listening on all channels
 
▶ CEO Agent · online · persistent memory loaded · 1,204 sessions
▶ Research Agent · online · FTS5 indexed · 160+ skills
▶ Builder Agent · online · Paperclip task queue connected
 
// Telegram DM from your phone will reach Hermes in <1s
$
0
$/month VPS cost
0
agents running Day 1
0+
skills per agent
1s
phone → agent latency

The Problem

You know AI agents are the future.
Nobody told you how to build one that actually works.

Every week there's a new agent framework, a new demo, a new promise. But between you and a real autonomous AI stack that runs 24/7, remembers everything, and talks to you from your phone — there's a wall of decisions nobody documents.

🤯

Which framework actually persists memory?

Most "agent" setups forget everything the moment the session ends. You're back to square one every time.

🏗️

How do you run agents 24/7 without breaking the bank?

Cloud AI services cost a fortune at scale. Nobody teaches you how to run this on a $5/month VPS.

📱

How do you actually talk to your agents from your phone?

Dashboards are fine. But you want to DM your AI OS like a person and have it do real work.

🔌

What's the control plane?

Raw agents have no governance. No cost controls, no org chart, no approval flows. It's chaos.

What's Inside

The complete playbook.
Every decision already made.

This isn't a theory doc. This is the exact architecture I built, validated against real source code, with every config, every command, and every gotcha documented.

01

VPS Selection + Bare Metal Setup

Hetzner CX22 at $5/month. Ubuntu 24.04 config, NVM, PM2 for auto-restart, Caddy for HTTPS in under an hour.

02

Hermes Agent Installation + Configuration

NousResearch's persistent memory agent (30+ tools, 80+ skills, SQLite across sessions, FTS5 search). The real alternative to stateless Claude Code.

03

Paperclip Control Plane

The open-source AI company OS. Visual org chart, heartbeat scheduler, cost controls, governance approvals — your agents' control plane.

04

Telegram Gateway — Your Phone Interface

Hermes Gateway setup for Telegram. DM your CEO agent from anywhere. <1 second latency. No laptop required.

05

Day 1 Org Chart — 3 Agents That Work

CEO (orchestration + memory), Research (Firecrawl + web intel), Builder (Claude Code + Paperclip tickets). Proven starting lineup.

06

Event-Driven Wakeups via n8n

Business signals (permit filings, competitor content, risk spikes) → n8n webhook → agent wakes with context. Proactive, not just scheduled.

07

VibeVoice Voice Layer

Speak to your agents hands-free on desktop. Microsoft's open-source VibeVoice-ASR via Vibing app — system-wide dictation, zero additional infrastructure.

08

Full Architecture Diagram + Cost Model

Every component mapped. Estimated cost: ~$69/month all-in for a full three-agent autonomous stack. Broken down line by line.

Why This Stack

vs. the alternatives

Feature Others Sovereign Stack
Persistent memory across sessions ✗ Stateless ✓ SQLite + FTS5
Phone access (no laptop) ✗ Dashboard only ✓ Telegram DM
Governance + cost controls ✗ Wild west ✓ Paperclip
Monthly infra cost $30–100+ $5/month
Voice interface ✗ None ✓ VibeVoice
Event-driven agent wakeups ✗ Cron only ✓ Webhook API
Skills library ✗ Build your own ✓ 160+ pre-built

Honest Comparison

Why not just use
Perplexity Computer?

Perplexity Computer is impressive. 19 AI models, 400+ integrations, $450M ARR in under 60 days. It proves the market is real. It also costs $200/month — and still can't do several things you actually need.

What matters Perplexity Computer Sovereign Stack
Monthly cost $200/mo minimum ~$69/mo all-in
Persistent memory across sessions ✗ Ephemeral cloud ✓ SQLite forever
Your data stays private ✗ Their servers ✓ Your VPS only
Phone access (Telegram DM) ✗ Web UI only ✓ Native
Cost transparency per task ✗ Opaque credits ✓ Exact $ per agent
Voice interface ✗ None ✓ VibeVoice
Model / provider lock-in ✗ Their routing ✓ 8 providers, your choice
You own the infrastructure ✗ Subscription dependency ✓ Always yours

Perplexity Computer is for people who want someone else to handle everything — and will pay $200/month for the convenience. The Sovereign Stack is for operators who want to own the infrastructure, control the costs, and keep the data. One weekend of setup saves you $1,572/year vs. Perplexity Max — and gives you capabilities they can't offer at any price.

Get It Now

One payment. Yours forever.

No subscription. No upsell. The full playbook — every config, every command, every gotcha — delivered as a PDF + architecture reference you keep.

The Sovereign Stack Blueprint
$97

One-time payment · Instant PDF download

  • ✓Complete VPS setup guide (Hetzner + PM2 + Caddy)
  • ✓Hermes agent install + configuration reference
  • ✓Paperclip control plane setup walkthrough
  • ✓Telegram Gateway — phone-to-agent in under an hour
  • ✓Day 1 org chart: 3-agent proven starting lineup
  • ✓n8n webhook → wakeup API integration guide
  • ✓VibeVoice voice layer setup (Mac)
  • ✓Full architecture diagram + annotated cost model
  • ✓Future updates included — free forever
Get Instant Access — $97

🔒 Secured by Gumroad · Instant PDF delivery · Questions? Reply to your receipt.

FAQ

Common questions

Do I need to be a developer to use this? ↓
You need to be comfortable with a terminal — copy-pasting commands, editing config files. You don't need to write code. If you've ever set up a Raspberry Pi or deployed a web app, this is well within reach. Every command is documented step by step.
What AI models does this work with? ↓
Hermes supports 8 LLM providers out of the box — Anthropic (Claude), OpenAI, Groq, Mistral, OpenRouter, and more. The playbook is built around Claude as the primary model, but you can swap providers without changing your agent configuration.
How long does the setup actually take? ↓
A focused weekend. Day 1: VPS provisioning + Hermes + Telegram gateway (4–6 hours). Day 2: Paperclip + first agents + n8n wakeup integration (4–6 hours). After that, your stack is running 24/7 with zero manual intervention.
Is this just a PDF or is there code included? ↓
It's a PDF playbook + architecture reference with every config file, PM2 ecosystem config, Caddy Caddyfile, Hermes gateway settings, and Paperclip adapter config included as copy-pasteable blocks. Nothing is hand-wavy — it's the exact setup.
What's the ongoing cost after setup? ↓
~$5/month for Hetzner CX22 VPS. Plus your API token usage (Claude, etc.) — with budget caps set in Paperclip, you control exactly what each agent is allowed to spend. Most users keep total monthly costs under $70 for a full three-agent stack with active usage.

Your AI OS is one weekend away.

Stop reading about autonomous agents. Build yours.

Get the Blueprint — $97

One-time. Instant download. Future updates free.