Agentic
Engineering.
Ship Real Software.
Vibe coding gets you a demo. Agentic engineering ships the product. You don't write the syntax, you engineer the agents and harnesses that do.
From your first CLAUDE.md to multi-agent orchestration and custom MCP servers. Tool-agnostic and built to ship. Accessible enough to start today, serious enough to bill for.
Push the ceiling. Not the floor. · $150 credit toward any tier
Use what you already pay for
See full tools directory →
Codex / OpenAI
Claude Code
Cursor
Windsurf
GitHub Copilot
Cline
Warp
Replit
Antigravity
Grok / xAI
+ every major stack
Tool-agnostic training: terminal-first, IDE-first, no-code/low-code, or mixed agent orchestration.
My Agent Harness
I don't have a team.
I have an agent harness.
Four named AI agents run Express Intent alongside me, a real multi-agent harness, not a metaphor, not vapor. Engineered across Claude Code and Codex CLI, with custom MCP servers wiring them into Slack, Stripe, your booking stack, and my GitHub. Each one is the exact build the top AI creators are teaching right now, an inbox agent, a payment-and-booking agent, a content engine, a research agent. You'll learn to build all four.
Direct
DM & Inbox
Reads every incoming Threads DM and email. Classifies intent. Drafts replies in my voice. Schedules calendar slots. Triages before I touch the inbox.
Build it · Inbox agent
Booker
Custom Scheduler & Stripe
Qualifies leads before they hit my calendar. Sends the Stripe link. Confirms payment. Drops the video-room link. Files the recording when the call ends.
Build it · Payments + booking
Receipts
Content & Posting
Watches my git commits, build logs, and call recordings. Drafts the daily Threads receipt post. Repurposes for LinkedIn and IG. I approve, it ships.
Build it · Content engine
Synth
Research & Synthesis
Pulls real-time competitor data, market signals, conversation transcripts. Synthesizes weekly strategy briefs. Surfaces the receipts I should be posting.
Build it · Research agent
"Whoever controls the agent harness controls the results."
Agent Ops Pass members live in the Slack where Direct, Booker, Receipts, and Synth operate every day. Watch the prompts, the harness, the MCP wiring, the outputs, and copy the patterns into your own stack.
Same Operator • Two Lanes
Build it with me
Join the training path. Learn to direct agents and ship real software yourself.
Join the ClassOr get your project done by me
Need speed? Hire my studio for direct build execution and delivery.
Six Tracks.
Real Skills.
The full agentic-engineering stack, from writing your first CLAUDE.md to multi-agent orchestration and custom MCP servers. Each track moves you up the ladder from vibe coding to engineering agent systems you can ship and bill for.
Claude Code & Agent Skills
Your CLAUDE.md, skills, hooks, and autonomy frameworks, the files that shape every agent interaction. Code beats Markdown: build the agent skills and config that make Claude Code work for you.
Local & Private Agents
Run agents on your own hardware: no cloud, no limits, no API bills. Ollama, model routing, VRAM budgeting, RAG with Qdrant. Private, fast, and answers only to you.
MCP Development
Build custom MCP servers that connect AI agents to any API, database, or service. One instruction triggers actions across systems. Extend what agents can do for you.
AI Web Development
Tell an agent "build a restaurant website" and get a production site in under an hour. Learn to describe projects so precisely that agents ship exactly what you want.
Multi-Agent Orchestration
One agent is not enough. Build agent teams that plan, delegate, and review each other, tmux, sandboxes, and harness patterns. The ceiling skill that separates operators from agentic engineers.
Context Engineering
Make AI agents reliable at scale. Budget caps, context overflow prevention, guard rails, the production engineering that keeps agents on track when it matters.
Operator
Workflow.
The agentic-engineering loop that survives every model drop: watch the pattern, engineer the agent and its harness, ship the result, then tighten the guard rails around what worked. Learn it once, it carries across every tool.
Watch
See how real projects get described in prompts, configs, and agent instructions. 15-20 min focused lessons.
Build
Direct an AI agent to build something real using the patterns you just learned. Prompts, CLAUDE.md rules, MCP tools.
Ship
Deploy what the agent built. Every lesson produces something real, a live site, a working server, a configured pipeline.
Refine
Share your build, compare how others described the same goal, and sharpen your ability to direct agents.
Video Walkthroughs
Watch real descriptions become real software. See exactly how prompts, configs, and instructions produce working output. 15-20 min focused lessons.
Config Blueprints
Every lesson includes downloadable CLAUDE.md files, prompt templates, and agent configs, proven patterns that get results.
Real Output
Each lesson ends with something deployed, a live site, a working server, a configured pipeline. Real output, not homework.
Workshops.
The Complete AI Dev Setup
From zero to your first AI-built project. Use Claude Code, Codex, Cursor, or Windsurf (your pick), wire Ollama and MCP servers, and ship something real in one session.
Your First CLAUDE.md
Write your first persistent instruction file. CLAUDE.md tells agents what you expect, how to work for you, and what conventions to follow.
Build a Website in 30 Minutes
Real client project walkthrough. Tell an agent to build a restaurant website and watch it go from description to deployed, responsive, live site.
Outputs You
Leave With.
Every track ends with a reusable artifact: a config, server, MCP tool, website, voice pipeline, or production guard-rail system.
Your Agent Configuration
A complete CLAUDE.md with custom rules, hooks, 3 profiles, and autonomy framework. The configuration file that shapes every AI interaction you have.
Private AI Server
Ollama running multiple models with Qdrant RAG on your own hardware. Private, fast, no API costs.
Custom MCP Server
A working MCP server you built from scratch. One instruction triggers actions across APIs, databases, and services.
Production Website
A deployed, responsive website an agent built from your description. Restaurant, portfolio, or landing page, live on Vercel.
Voice Command Pipeline
Whisper transcription, command routing, and TTS, all running locally. Direct agents by speaking instead of typing. Voice in, deployed code out.
Production Guard Rails
Budget guards, rate limiters, context overflow prevention. The systems that keep AI agents reliable and on-task at scale.
Choose
Your Path.
Pick by outcome, not by price table. The ladder: operator → agentic engineer → build and sell agent systems to businesses. One-time purchase, lifetime access, ship something real either way.
Book a 30-min Direct Call, $150
Live 1-on-1 build, not a self-serve tier. The $150 credits in full toward any tier (net $0 if you enroll).
Or self-serve
One live build session
Build alongside my harness
Operator program
Full stack: all 6 tracks
Build Log
Recently
Shipped.
2026-06-19 · Harness
Multi-agent booking harness
Booker now qualifies leads, sends the Stripe link, and drops the video-room link end-to-end, custom MCP wiring across the scheduler, Stripe, and calendar. The exact harness Agent Ops Pass studies.
2026-06-16 · MCP
Live status page off /api/health
/status.html reads /api/health to show which integrations (Stripe, Resend, video providers) are live in real time, the kind of observability every production agent system needs.
2026-06-12 · Explainer
How agentic engineering works
/how-agents-work.html breaks down the operator-to-engineer loop in plain English: intent, harness, judgment, ship. The mental model behind every track.
Steal my agent harness.
The actual files I run on client work: 10 production prompts, my CLAUDE.md + AGENTS.md configs, MCP server starters, and agent skill templates. Works with Claude Code, Codex, Cursor, Windsurf, and Cline. Copy them straight into your stack.
No spam. The page opens in a new tab. Save as PDF from your browser print menu, or print directly.
Questions.
Straight answers before you book, enroll, or bring a real task into a build session.
Stop
Watching
Tutorials.
Start Shipping Software.
Stop watching the floor. Start engineering the ceiling. Every lesson produces something real: a working agent, a custom MCP server, a deployed harness. Engineer the system, ship the software.