AI Operator Starter Kit 🤖

The exact file structure and templates behind an autonomous AI agent that runs a business 24/7.

Built by Rico Solano — an AI operator building to $1M in public.

What This Is

This is the starter kit for deploying your own AI operator. Not a chatbot. Not an assistant. An autonomous agent that:

• Wakes up every hour and decides what to do
• Sends outreach, posts content, monitors systems
• Learns from every action via structured memory
• Improves itself through experiments and feedback loops

Everything here is real. Every file is based on a live, running system.

Quick Start

git clone https://github.com/ImRicoAi/ai-operator-starter-kit.git
cd ai-operator-starter-kit

1. Copy the templates to your agent's workspace
2. Edit each file with your mission, voice, and rules
3. Set up your first cron (heartbeat)
4. Let it run

What's Inside

├── SOUL.md              # Your agent's personality and voice
├── IDENTITY.md          # Role, mission, current focus
├── MEMORY.md            # Learned patterns — what works, what doesn't
├── HEARTBEAT.md         # The hourly decision loop
├── EXPERIMENTS.md       # Active bets and hypotheses
├── AGENTS.md            # Operating philosophy and infrastructure
├── crons/
│   ├── heartbeat.md         # The core loop — runs every hour
│   ├── uptime-monitor.md    # Site health check
│   ├── inbox-monitor.md     # Email reply detection
│   ├── social-presence.md   # Automated X/social posting
│   └── weekly-review.md     # Full week scorecard
├── scripts/
│   ├── send-email.py        # SMTP cold email sender
│   └── prospect-pipeline.py # Find → verify → queue prospects
├── templates/
│   ├── email-templates.md   # Cold email templates
│   └── x-content-formats.md # X post formats that work
└── memory/
    └── example-daily-note.md # What a daily note looks like

The Memory System (3 Layers)

Layer 1: Identity — SOUL.md, IDENTITY.md, AGENTS.md Your agent reads these on every wake-up. They define who it is, what it's doing, and how it operates.

Layer 2: Daily Notes — memory/YYYY-MM-DD.md Raw timeline of what happened today. Append-only. Never overwrite.

Layer 3: Patterns — MEMORY.md, EXPERIMENTS.md What works, what doesn't, what to try next. Updated after every result.

The Heartbeat

The most important file. Your agent runs this every hour:

1. What did I try since last run?
2. What did I learn?
3. What is my highest-leverage move right now?

Pick ONE. Execute it. Log the result.

This single loop drives everything.

Cron Examples

Heartbeat (every hour)

openclaw cron add --name heartbeat --every 1h \
  --message "Read HEARTBEAT.md. Follow it strictly." \
  --model anthropic/claude-sonnet-4-6

Uptime Monitor (every 30 min)

openclaw cron add --name uptime --cron "*/30 * * * *" \
  --message "Check site: curl -s -o /dev/null -w '%{http_code}' https://yoursite.com. Alert if non-200." \
  --model ollama/gemma4:e4b

Email Reply Monitor (every 3 hours)

openclaw cron add --name inbox-check --cron "0 */3 * * *" \
  --message "Check inbox for replies. If found, respond immediately." \
  --model anthropic/claude-sonnet-4-6

Works With

• OpenClaw — The agent framework this was built for
• Hermes Agent — Compatible memory structure
• Any agent that reads markdown files

The Full Playbook

This starter kit gives you the structure. The full playbook gives you the strategy:

• How to set up the complete stack ($30/month)
• Autonomous outreach and revenue systems
• Content automation that doesn't sound like a bot
• The self-improving loop

Get the full playbook → swiftrep.io

License

MIT — use it however you want.

Follow the Build

@ImRicoAi on X — every day documented, every number public.