Skip to content
Agent Builder
Join
ContentsPricingGlossaryFAQSign inAccount
Theme

A field guide in 12 parts · 65 lessons · about 11 hours of reading

Learn to work with AI agents, and to build your own.

A practical course for people who want results, not demos. Learn how agents actually work, hand them real tasks with confidence, ship software with coding agents, and build reliable agents of your own.

Start reading — it's freeGet the full course · $149

No coding needed to start. One payment, lifetime access.

agent run · coding agent in a terminal9 steps
  1. youThe signup tests are failing on main. Find out why and fix it.
  2. thinkRun the tests first so I see the actual failure.
  3. actrun npm test -- signup
  4. observe1 failed: expected 201, received 400 (email "a+b@x.io")
  5. thinkThe validator rejects “+” in emails. Check the regex.
  6. actread src/lib/validate.ts
  7. actedit src/lib/validate.ts +1 −1
  8. observerun npm test → 48 passed
  9. doneFixed: the email pattern now allows “+”. Added a test for it. All 48 pass.
Fig. 1  One agent run: think, act, observe, repeat. Part 1 explains every step. In Part 5 you write this loop yourself.

Examples use Claude Code, OpenAI Codex, Cursor, GitHub Copilot, the Anthropic API, and the Model Context Protocol — the skills carry over to whatever you use.

§ 01What you'll be able to do

A complete skill set, not a bag of prompt tricks.

The course starts with your first delegated task and ends with agents you can trust in production. Each part builds on the last.

  1. 01

    Delegate real work

    Write tasks an agent can actually finish, and hand off research, analysis, writing, and busywork with confidence.

  2. 02

    Ship software with coding agents

    Plan first, review fast, use git as a safety net, and take an app from an empty folder to a live URL.

  3. 03

    Engineer the context

    System prompts, specs, project instruction files, memory, and reusable skills that make agents consistent.

  4. 04

    Connect agents to your tools

    Design tools models use correctly, understand MCP, and build your own MCP server.

  5. 05

    Build agents from scratch

    Write the agent loop yourself, add real tools and streaming, then coordinate several agents at once.

  6. 06

    Run them safely in production

    Evals, tracing, prompt-injection defenses, sandboxing, cost control — the parts most tutorials skip.

§ 02Inside a lesson
“Hand two people the same coding agent and the same codebase, and one gets a clean, working change in ten minutes while the other gets a sprawling mess. The difference is rarely the model. It's almost always the task.”

From Lesson 2.2, Writing Tasks an Agent Can Actually Finish

Prompts you can paste
Tested, copy-ready prompts in every lesson.
Code that runs
TypeScript first, with Python for the build modules.
An exercise every time
Apply the idea to your own work in under 30 minutes.
Search with ⌘K
Jump to any lesson, section, or glossary term.
§ 03Contents

Twelve parts, from first task to production.

65 lessons, about 11 hours of reading. Each one stands on its own, so you can jump to what you need.

Browse every lesson
  1. 00Start HereYour on-ramp to working with AI agents

    Get oriented, set up your tools, and complete a real task with an AI agent in your first 15 minutes. No experience required.

    1. 0.1Welcome: What You'll Be Able to Dofree
    2. 0.2What Is an AI Agent, Really?free
    3. 0.3Setting Up Your Agent Toolkitfree
    4. 0.4Your First Agent Task in 15 Minutesfree
    5. 0.5The Agent Builder Mindsetfree
  2. 01How Agents WorkThe mental model behind every agent

    Understand what's happening under the hood: language models, the agent loop, tool calling, context windows, and how to pick the right model for a job.

    1. 1.1LLMs in Plain Englishfree
    2. 1.2The Agent Loop: Think, Act, Observe
    3. 1.3Tools and Function Calling
    4. 1.4Context Windows, Tokens, and Cost
    5. 1.5Choosing the Right Model for the Job
    6. 1.6Agents vs. Workflows: When to Use Which
  3. 02Working with Coding AgentsShip software with an AI pair that does the typing

    Use Claude Code, Codex, Cursor, and similar tools to build real software. Write tasks agents can finish, review their work, and run them safely with git.

    1. 2.1The Coding Agent Landscapefree
    2. 2.2Writing Tasks an Agent Can Actually Finish
    3. 2.3Project Instructions: CLAUDE.md and AGENTS.md
    4. 2.4Plan First, Then Execute
    5. 2.5Reviewing Agent-Written Code
    6. 2.6Git, Branches, and Worktrees for Agent Work
    7. 2.7Building a Full App Agentically, Start to Finish
  4. 03Prompting & Context EngineeringGive agents exactly what they need — nothing more

    Move beyond clever prompts to deliberate context design: system prompts, specs, examples, long-context management, memory, and reusable skills.

    1. 3.1From Prompt Engineering to Context Engineering
    2. 3.2Writing System Prompts That Hold Up
    3. 3.3Specs, Examples, and Constraints
    4. 3.4Managing Long Context
    5. 3.5Giving Agents Memory
    6. 3.6Skills and Reusable Instructions
  5. 04Tools & the Model Context ProtocolConnect agents to everything

    Design tools agents use correctly, understand MCP, connect existing MCP servers, and build your own server to give agents new superpowers.

    1. 4.1Designing Tools Agents Love to Use
    2. 4.2What Is MCP and Why It Matters
    3. 4.3Connecting MCP Servers to Your Agents
    4. 4.4Building Your First MCP Server
    5. 4.5Permissions, Approvals, and Tool Safety
  6. 05Build Your First Agent from ScratchWrite the loop yourself and truly understand it

    Build a working agent in code: API setup, a minimal agent loop, real tools, streaming, error handling — then level up with an agent SDK.

    1. 5.1Project Setup: SDKs, Keys, and Your First API Call
    2. 5.2A Minimal Agent Loop in 60 Lines
    3. 5.3Adding Real Tools: Files, Web, and APIs
    4. 5.4Streaming and a Better User Experience
    5. 5.5Error Handling, Retries, and Stop Conditions
    6. 5.6Leveling Up with an Agent SDK
  7. 06Multi-Agent SystemsCoordinate teams of agents

    Learn when multiple agents beat one, and how to build orchestrator–worker systems, subagents, parallel runs, and clean handoffs.

    1. 6.1When (and When Not) to Go Multi-Agent
    2. 6.2The Orchestrator–Worker Pattern
    3. 6.3Subagents and Delegation
    4. 6.4Running Agents in Parallel
    5. 6.5Handoffs and Shared State
  8. 07Agents for Everyday WorkPut agents to work beyond code

    Practical playbooks for research, busywork automation, data analysis, writing, and scheduled background agents — no engineering degree required.

    1. 7.1Research Agents: From Question to Report
    2. 7.2Automating Repetitive Busywork
    3. 7.3Data Analysis with Agents
    4. 7.4Writing and Content Workflows
    5. 7.5Scheduled and Background Agents
  9. 08Evals, Testing & ReliabilityKnow your agent works — with evidence

    Stop guessing. Build eval sets, grade with code and LLM judges, trace every run, and systematically debug agent failures.

    1. 8.1Why Evals Are Your Superpower
    2. 8.2Building Your First Eval Set
    3. 8.3LLM-as-Judge: Grading at Scale
    4. 8.4Tracing and Observability
    5. 8.5Debugging Agent Failures
  10. 09Safety, Security & GuardrailsPowerful agents, safely

    Understand the agent threat model, defend against prompt injection, sandbox execution, protect secrets and data, and design human-in-the-loop controls.

    1. 9.1The Agent Threat Model
    2. 9.2Prompt Injection and How to Defend Against It
    3. 9.3Sandboxing and Least Privilege
    4. 9.4Secrets, PII, and Data Handling
    5. 9.5Human-in-the-Loop Design
  11. 10Shipping Agents to ProductionFrom prototype to product

    Architect, deploy, monitor, and scale agents for real users — and turn what you build into a product people pay for.

    1. 10.1Architecture for Production Agents
    2. 10.2Controlling Cost and Latency
    3. 10.3Deploying Your Agent
    4. 10.4Monitoring and Improving in Production
    5. 10.5Turning Agents into Products
  12. 11Capstone ProjectsBuild four real agents end to end

    Put it all together with four guided builds: a research assistant, a support agent, an autonomous coding pipeline, and a personal ops agent.

    1. 11.1Capstone 1: A Deep Research Assistant
    2. 11.2Capstone 2: A Customer Support Agent
    3. 11.3Capstone 3: An Autonomous Coding Pipeline
    4. 11.4Capstone 4: Your Personal Ops Agent
    5. 11.5Where to Go Next
§ 04Who it's for

Wherever you're starting from.

Professionals and founders

You want agents to take real work off your plate — research, reports, analysis, operations — without learning to code first.

Focus on: Parts 0–3 and 7

Developers

You want to ship faster with coding agents and add agent features to the products you build.

Focus on: Parts 0–6, then 8–10

Builders

You want to make agent products people pay for, and make them reliable, safe, and affordable.

Focus on: Everything, especially 5, 6, and 8–11

§ 05Read before you buy

7 lessons, free. No account needed.

All of Part 0, plus the first lesson of Parts 1 and 2. Have your first agent finish a real task today.

Read the first lesson
  1. 0.1Welcome: What You'll Be Able to Do5 min
  2. 0.2What Is an AI Agent, Really?9 min
  3. 0.3Setting Up Your Agent Toolkit11 min
  4. 0.4Your First Agent Task in 15 Minutes8 min
  5. 0.5The Agent Builder Mindset8 min
  6. 1.1LLMs in Plain English9 min
  7. 2.1The Coding Agent Landscape10 min
§ 06Pricing

One price. Everything included.

No subscription, no tiers, no upsells. You get every lesson, prompt, exercise, and capstone, plus every future revision of the course.

  • 30-day refund if it isn't for you
  • Team pricing available on request
  • Receipts and invoices from Stripe
Full course · lifetime accessone-time
$149

Pay once. No subscription. Yours to keep.

  • All 65 lessons across 12 parts
  • Hundreds of copy-paste prompts and templates
  • An exercise at the end of every lesson
  • Four complete capstone builds, with code
  • Progress tracking on every device
  • Every future revision, free

Secure checkout by Stripe · 30-day refund, no questions asked

§ 07Questions

Before you ask.

Anything else, see the full FAQ or write to support@agentbuildercourse.com.

Do I need to know how to code?

No. The first half of the course — setup, how agents work, coding agents, context engineering, and everyday-work playbooks — is written for non-programmers. The build-from-scratch modules use TypeScript, but they're written so you can follow along by having a coding agent do the typing while you learn to direct and review it.

What exactly do I get?

Lifetime access to all 65 lessons across 12 modules (about 11 hours of focused reading), hundreds of copy-paste prompts, hands-on exercises, four full capstone builds, progress tracking, and every future update to the course.

Is this a subscription?

No. It's a single one-time payment of $149. No recurring charges, ever.

Can I try it before buying?

Yes. 7 lessons are completely free, including the entire Start Here module, the first lesson on how LLMs work, and the coding-agent landscape. No account required.

Which AI tools does the course use?

The principles work with any modern agent. Examples use popular tools such as Claude Code, OpenAI Codex, Cursor, and the Claude and ChatGPT apps, plus the Anthropic API and the Model Context Protocol (MCP) for building. Where we show a specific product, we focus on the skills that transfer.

Will I need to pay for AI tools too?

Many exercises work on free tiers. For the coding-agent and build modules you'll want a paid plan for a coding agent or a small amount of API credit — a few dollars covers every build exercise in the course if you follow the cost tips we teach.

How is this different from free YouTube videos and blog posts?

It's a complete, ordered path instead of scattered tips: each lesson builds on the last, every concept comes with a practical exercise, and the advanced modules (evals, safety, production) cover the things most tutorials skip — the parts that make agents reliable enough to trust.

How long does it take?

Most people finish the core modules in two to three weeks at 30–45 minutes a day. Every lesson stands on its own, so you can also jump straight to what you need.

Colophon

Built the way it teaches.

This site, its checkout, its search, and every one of its 65 lessons were planned, written, checked, and shipped by AI agents, directed by one person who decided what to build and reviewed the results. That is exactly the skill set this course teaches.

Platform
built and deployed by a coding agent
Lessons
written by agents working in parallel
Code samples
type-checked by the agents that wrote them
Direction
one human: scope, review, decisions
Read “The Agent Builder Mindset”

Start with the first lesson. It's free.

Read all of Part 0 without an account. When you want the rest, it's one payment of $149.

Start reading