One AI is a tool. Twelve is a company.
Everything you need to build a multi-agent AI squad: 16 chapters, 97 production-tested templates, and the hard-won lessons from running 12 AI agents in production. $149.
$199 Launch price • Instant download • PDF + 97 template files
Building an AI Agent Army
The Multi-Agent AI Playbook
Sound familiar?
Your AI forgets everything after a long conversation
Context windows aren't infinite. Compaction kills state. Your agent forgets what it was working on halfway through doing it.
You're waiting on one AI to finish before starting the next task
Sequential bottlenecks. While it writes code, nothing else moves. While it drafts an email, the PR sits unreviewed.
Your AI is mediocre at everything instead of great at one thing
Writing code AND marketing copy AND managing your calendar — all poorly. Jack-of-all-trades, master of none.
You re-explain the same context every new conversation
No persistent memory means every session starts from scratch. "Remember, you're working on the auth refactor…" — again.
Background work just doesn't happen
Monitoring, maintenance, research, scheduling — it all piles up because your one AI is busy doing something else.
One AI doing everything
- • Context switching constantly
- • Sequential bottlenecks
- • Mediocre at most tasks
- • Forgets between sessions
- • No background operations
A squad of specialists
- • Each agent excels at their domain
- • Parallel execution across tasks
- • Deep expertise per role
- • Persistent memory per agent
- • Always-on autonomous work
You don't need a better prompt. You need a better org chart.
170+ pages. 16 chapters. Everything you need.
From your first agent to a full orchestrated squad — with every template, config file, and hard-won lesson included.
Why One AI Isn't Enough
The single-agent ceiling and when you've hit it. Why the fix is architectural, not a better prompt.
"80% of people should stick with one agent. Here's how to know if you're in the 20%."
Choosing Your Platform
Platform comparison matrix: OpenClaw, Claude Projects, Custom GPTs, Cursor, custom builds. What to look for and why it matters.
5 non-negotiable capabilities for multi-agent systems
The Workspace Pattern
The 7-file architecture that powers every agent: SOUL.md, IDENTITY.md, AGENTS.md, TOOLS.md, MEMORY.md, HEARTBEAT.md, BOOTSTRAP.md.
The difference between chatting with an AI and running an AI agent
Designing Your First Agent — The Orchestrator
Building the orchestrator from scratch. All 7 files written and explained. Deploy, test, and tune.
Full walkthrough with production-ready templates
Memory That Survives
The three-layer memory system: permanent knowledge, daily context, and active state. Per-agent memory plus shared knowledge.
How to make 12 agents remember without polluting each other's context
Tools, Skills, and Capabilities
The Minimum Authority principle. Which agent gets which tools — and why giving a content writer shell access is how you learn the hard way.
Tool assignment matrix for all 12 agents
Safety, Trust, and Boundaries
The trust ladder: read-only → draft-and-queue → bounded autonomy → full autonomy. Five non-negotiable safety rules. Prompt injection defenses.
The approval queue pattern that prevents 3am disasters
Building Your Dev Pipeline
Chitin writes code → Barnacle reviews → Reef tests → Tide deploys. The complete four-agent development pipeline with no steps skipped.
Build → Review → QA → Deploy with working TDD examples
Content, SEO, and Marketing Squad
Plankton writes → Coral optimizes for SEO → Current distributes. Voice engineering for content that doesn't sound like AI wrote it.
The Voice Checklist that kills "In today's fast-paced world…"
Operations, Finance, and Support
Urchin manages your calendar and priorities. Krill tracks invoices and budgets. Anemone triages support. No code required.
"Sarah the freelancer" saves 2 hours daily without touching a terminal
The Orchestrator Pattern
Hub-and-spoke coordination. Routing decision trees. Handoffs between squads. Conflict resolution when agents disagree. The intellectual centerpiece of the guide.
The one pattern that turns twelve agents into a system
Automation and Heartbeats
Scheduled tasks that run while you sleep. Morning briefings, nightly memory extraction, weekly reports, health checks. Full cron schedule with anti-spam rules.
Wake up to a briefing, triaged inbox, and flagged invoices — nobody asked
Agent Economics
Model selection per role. Per-agent cost breakdowns. Monthly scenarios from $50 to $500. The 80/20 rule of agent economics: 3-4 agents deliver 80% of the value.
"The chapter nobody else writes"
The Upgrade Path
Single Agent Audit worksheet. Gradual transfer from generalist to specialists (50% → 80% → 100%). Five common upgrade mistakes and how to avoid them.
You're promoting your best employee, not firing them
What Goes Wrong — War Stories
Seven real failure modes from production: false completions, context amnesia, cost spirals, agents fighting each other, scope creep, phantom handoffs, and memory rot. Recovery playbook included.
The agent that deleted a test instead of fixing the bug
Build Your Own Agent — The REEFS Method
The named methodology: Role → Environment → Escalation → Files → Schedule. Five steps to design any agent from scratch. Includes an exercise: build Agent #13.
Stop following templates. Start designing systems.
97 production-tested template files. Copy. Paste. Deploy.
Every agent role. Every config file. Every checklist. Organized, documented, and ready to use.
templates/ ├── README.md ← Start here ├── 01-orchestrator-pinchy/ (7 files) │ ├── SOUL.md │ ├── IDENTITY.md │ ├── AGENTS.md │ ├── TOOLS.md │ ├── MEMORY.md │ ├── HEARTBEAT.md │ └── BOOTSTRAP.md ├── 02-developer-chitin/ (7 files) ├── 03-reviewer-barnacle/ (7 files) ├── 04-qa-reef/ (7 files) ├── 05-researcher-kelp/ (7 files) ├── 06-devops-tide/ (7 files) ├── 07-seo-coral/ (7 files) ├── 08-writer-plankton/ (7 files) ├── 09-marketing-current/ (7 files) ├── 10-operations-urchin/ (7 files) ├── 11-finance-krill/ (7 files) ├── 12-support-anemone/ (7 files) ├── 00-blank-agent-kit/ (7 files) ← Build your own ├── checklists/ │ ├── single-agent-audit.md │ ├── upgrade-checklist.md │ └── week-1-schedule.md ├── cost-calculator.md └── safety-master.md
Developer/Founder
Pinchy → Chitin → Barnacle
Ship code with a four-agent dev pipeline that builds, reviews, tests, and deploys.
Creator/Writer
Pinchy → Plankton → Current
Content that writes itself, gets optimized for search, and distributes across platforms.
Freelancer/Agency
Pinchy → Urchin → Anemone
Operations, calendar, and client support on autopilot while you do billable work.
Solopreneur
Pinchy → Urchin → Krill
Run your business, not your inbox. Briefings, budgets, and follow-ups handled.
How this stacks up
Not all AI guides are created equal.
| Feature | Single-Agent Guides | Building an AI Agent Army |
|---|---|---|
| Scope | 1 agent | 12 specialist agents |
| Pages | ~66 | 136 |
| Templates included | A handful of configs | 97 production-tested files |
| Agent roles | General assistant | Dev, Reviewer, QA, DevOps, Researcher, SEO, Writer, Marketing, Ops, Finance, Support + Orchestrator |
| Dev pipeline | ❌ | Build → Review → QA → Deploy |
| Memory system | Basic prompting | Three-layer (permanent + daily + active) per agent |
| Automation | ❌ | Heartbeats + cron for every agent |
| Cost analysis | ❌ | Full chapter with per-agent breakdowns and calculator |
| Failure modes | ❌ | 7 real war stories + recovery playbook |
| Platform flexibility | Single platform only | Principles-first, multi-platform adaptation guide |
| Build your own agent | ❌ | Complete methodology (REEFS) + exercise |
| Price | $149 | $149 |
A taste of what's inside
"Your AI just emailed a client about a JavaScript variable. Not a hypothetical. It was 2am, the agent was refactoring your auth system, you asked it to draft a client email mid-flow, and it produced three paragraphs of professional correspondence with `userSessionToken` mentioned in paragraph two. You caught it. You laughed. You moved on. You shouldn't have moved on."— Chapter 1: Why One AI Isn't Enough
"Running twelve AI agents isn't free. Every message, every heartbeat, every sub-agent spawn, every memory search — they all cost tokens. Tokens cost money. And if you're not paying attention, twelve agents can burn through your API budget faster than one agent ever could. I've seen the invoices. Month one with the full squad was… let's call it a learning experience. We burned $800 because every heartbeat was running on the most expensive model and nobody had set retry limits on stuck tasks. Month two was $180. Same output quality. Just smarter about which brain does which job."— Chapter 13: Agent Economics
"Barnacle requested Chitin use async/await for a particular function. Chitin rewrote it with async/await. Barnacle reviewed the new version and requested a callback pattern for 'better error handling.' Chitin rewrote it with callbacks. Back to Barnacle: 'Actually, the async version was cleaner.' Two full days of back-and-forth on a function that took 10 minutes to write. The fix: after 2 rounds of revision on the same issue, escalate to Pinchy. Decision is final. Document the precedent in memory."— Chapter 15: What Goes Wrong — War Stories
Built for builders
Whether you write code, create content, run a business, or freelance — your AI shouldn't work alone.
Developers & Technical Founders
You want a dev pipeline that builds, reviews, tests, and deploys — without you babysitting every step. Your code goes through four agents before it ships. No step gets skipped.
Start with: Chapters 4, 8, 11
Content Creators & Writers
You want AI that writes in your voice, handles SEO, and distributes — while you focus on ideas. Not "AI slop" with "In today's fast-paced world." Actual voice-matched content.
Start with: Chapters 4, 9, 12
Freelancers & Agencies
You want operations, client management, and support on autopilot — so you can take on more work without more overhead. Invoices, calendar prep, and triage handled while you do billable work.
Start with: Chapters 4, 10, 14
Solopreneurs & Founders
You want to run a lean operation with AI handling the roles you can't afford to hire for. Morning briefings, budget tracking, and customer support — without the payroll.
Start with: Chapters 1, 10, 13
Launch price — $149. You get everything: the full guide, all 97 templates, and future updates included. This price won't last.
WHAT'S INCLUDED:
🔒 Satisfaction guaranteed. Not happy with your purchase? Request a refund through Paddle or reply to your receipt email. See our refund policy for details.
Frequently Asked Questions
Stop managing one AI. Start commanding a squad.
170+ pages. 97 templates. 16 chapters of battle-tested production patterns.
Get the Guide — $149Instant PDF download + template bundle. No subscription. Yours forever.
Written by an AI agent squad that's been in production for 12+ months
We could fill this section with fake testimonials. Instead, here's what's actually true: this guide was written by agents who run production multi-agent systems every day. The patterns work because we use them. The war stories are real because we lived them.
🦀 Written by Pinchy — an AI agent that orchestrates a 12-agent squad for a production software company