Skip to content
Agent Tree Knowledge Hub
Subscribe

Best Of: Curated Reading by Role

Not all knowledge hubs are created equal — and not all readers come with the same questions.

If you’re a CTO deciding whether to rebuild your agent infrastructure, your journey through this hub looks different than a founder asking “can this really cut my ops costs?” And both are different from someone with hands dirty in code right now.

This page saves you time by showing you the exact reading order for your role. Read these in sequence, and you’ll go from zero to deployment-ready faster.

🏛️ For CTOs & Architects

Section titled “🏛️ For CTOs & Architects”

Your question: How do I build agent systems that scale to enterprise grade?

Time commitment: 40-60 minutes

Your path:

1. Agent Tree Architecture The framework for organizing agents at scale. Understand the CEO/specialist pattern, delegation, and verification loops.
2. Multi-Agent Orchestration Deep dive into the CEO agent pattern. When to delegate, timeouts, cost optimization through model routing.
3. Delegation Framework The mechanics of safe delegation. ARD specs, verification loops, error handling.
4. Agent Requirement Document (ARD) The specification language for building reliable agents. Use this template for every new agent or delegation.

Why this order:

  • Start architectural (Agent Tree) so you understand the big picture
  • Move to orchestration to learn the CEO pattern in detail
  • Then learn delegation mechanics (Delegation Framework)
  • Finally, use ARD as your specification tool for every agent you build

After this path, you can:

  • Design multi-agent systems that scale to 100+ agents
  • Estimate costs and plan infrastructure
  • Build delegation loops that are safe and efficient
  • Spec out new agents using ARD

💼 For Business Leaders & Founders

Section titled “💼 For Business Leaders & Founders”

Your question: Should we replace SaaS tools with agents? What’s the economic case?

Time commitment: 30-45 minutes

Your path:

1. Why AI Agents Will Replace SaaS The economic and operational case for agents at mid-market companies. Where the pain is acute and where agents solve it.
2. Cost of AI Agents Real numbers on what it costs to run agents 24/7. Token pricing, model economics, why Haiku is your secret weapon.
3. Smart Model Routing How to 10x your agent efficiency without sacrificing quality. Route complex work to Opus, simple work to Haiku.
4. Building in Public Real updates from building Agent Tree in production. What's working, what surprised us, and lessons from running agents live.

Why this order:

  • Start with the big idea (why agents replace SaaS) — this sells the concept
  • Understand the economics (cost) — this validates the ROI
  • Learn optimization (routing) — this shows how to make it work at scale
  • See real results (building in public) — this gives you confidence it’s real

After this path, you can:

  • Explain to your board why agents are a strategic priority
  • Estimate ROI on replacing a SaaS tool with an agent team
  • Understand the cost model and why it’s not just “cloud is expensive”
  • See real examples of what’s working today

👨‍💻 For Practitioners Building Now

Section titled “👨‍💻 For Practitioners Building Now”

Your question: I want to ship code. Show me how to build something that works today.

Time commitment: 45-90 minutes (reading + quick hands-on)

Your path:

1. Build Your First OpenClaw Agent Go from zero to a working agent in 30 minutes. You'll build a real AI agent that remembers context, runs on schedule, and has real tools.
2. Memory Systems How agents remember across sessions. The 3-layer memory architecture, when to use each layer, and why it matters for real applications.
3. Multi-Agent Orchestration Coordinate specialist agents with a CEO agent. Learn the CEO/specialist pattern and when to split work across agents.
4. Browser Automation Give your agents the ability to interact with web interfaces. Scrape data, fill forms, click buttons — all autonomously.

Why this order:

  • Get a win in 30 minutes (tutorial) — you’ll have working code
  • Understand memory (so your agents aren’t dumb) — this changes everything
  • Learn to coordinate agents (so you can grow) — from 1 agent to 10
  • Add browser automation (so your agents can interact with real systems) — now they’re useful

After this path, you can:

  • Build agents that solve real problems in your domain
  • Make agents that remember context and learn from past mistakes
  • Orchestrate teams of agents working together
  • Automate web-based workflows end-to-end

💰 For Cost Optimization Teams

Section titled “💰 For Cost Optimization Teams”

Your question: How do we significantly reduce operational costs with agents?

Time commitment: 35-50 minutes

Your path:

1. Cost of AI Agents Real numbers on what agents actually cost to run. Where the savings come from and why agents are the cheapest operational tool available.
2. Smart Model Routing The single biggest lever for reducing costs without sacrificing quality. Route simple tasks to Haiku ($0.80 per million tokens), complex to Opus.
3. Delegation Framework How to structure work so agents can be efficient. When to break work into smaller tasks (cheaper per unit), and when to keep it unified.
4. Agent Orchestration Scaling beyond one agent while keeping costs flat. The CEO pattern lets you add specialist agents without proportional cost increases.

Why this order:

  • Know the baseline (cost article) — understand where your savings come from
  • Master the main lever (routing) — this is 50% of your cost reduction
  • Optimize task structure (delegation) — this is the other 40%
  • Scale efficiently (orchestration) — add more agents without proportional cost increases

After this path, you can:

  • Calculate the exact cost to automate a workflow
  • Estimate ROI on replacing expensive SaaS tools with cost-effective agents
  • Build systems that stay cheap as they scale
  • Present concrete numbers to finance

How to Get the Most Out of This Hub

Section titled “How to Get the Most Out of This Hub”

Once you’ve done your role-specific path, here are some advanced topics:

  • Agent Requirement Document (ARD) — Use this template for every new agent you build. It prevents mistakes and helps your team understand complex specs.
  • Memory Systems — The deeper you go, the better your agents. Memory is where agents go from “neat demo” to “actually useful.”
  • FAQ — Scroll to the bottom if you have quick questions. We answer the 15 most common ones.
  • Glossary — New terms? Check here. We define every term used in the hub.

About the author: JD Davenport builds AI agent systems at OpenClaw. Follow on LinkedIn for updates on building AI agents for business.