Prototype AI agents the way you prototype apps. In plain text, no code.
A practical live online training where you build, test, and run production ready AI agents inside a live browser environment.
role: Competitive analysis agent
tools: web search, pricing scraper, docs reader
guardrails: none
Super Early Bird · first 30 seats
Instant access receipt · Supports Apple Pay, Google Pay & Corporate Cards
✓ 100% Week 1 Guarantee: Run live agents in Session 1 or get a full 1-click refund.
- Run your first live agent in minute 15
- 100% browser-based lab (zero local setup or IT tickets)
- Includes 1-click corporate L&D reimbursement receipt
How it works
Define it.
Write how your agent should behave in plain text using Raw Agent Control (RAC).
Run it and break it.
Run it in your browser on real scenarios, catch it the moment it breaks, and update your definitions until it runs smoothly.
Hand it off.
Give your engineers, Cursor, or Claude Code a tested agent spec they can build on and check their version against.
What you'll walk away with
Shape behavior without waiting on engineering
You'll be able to define an agent's role, tools, memory and guardrails in plain text, so you can shape its behavior without waiting on engineering.
Catch failures before users do
You'll be able to spot the four ways agents break and fix each one in the text, so problems surface before your users find them.
Ship what you actually designed
You'll be able to hand engineering a tested agent and its run traces, so what ships matches what you designed.
The four weeks
Agent brains.
Run a live agent in the first 15 minutes, then define its role, context, decisions and output.
Tools and memory.
Connect tools and MCP, control what the agent remembers, and pass work between agents.
When agents fail.
Trigger and fix endless loops, forgotten rules, made up data and errors that cascade.
Production control.
Add guardrails, automated quality checks and human sign off, then package the spec for engineering.
Do I have to learn to code, or master yet another platform?
Neither. You define your agent in plain text, run it live on real scenarios, and hand it straight to your engineers. Nothing to install, nothing to get locked into.
Isn't this just another demo that breaks in production?
That's exactly what you'll practice. You throw real scenarios at your agent, watch it fail, and fix it before it ever reaches users.
Isn't this just another PRD?
No. It's what Lovable did for apps, applied to agents. Instead of a document engineers have to interpret, you hand them an agent you've already run live on real data and real scenarios. They can follow its traces step by step and check their production version against it.
The details
- Live on Zoom once a week, Wednesdays at 12:00 PM EST / 7:00 PM IST
- Every session is recorded, and each week starts with a live build recap so even if work makes you miss a week, you never fall behind
- Runs in your browser, nothing to install
- First cohort starts November 18, 2026
- Taught live by Yigal, CTO, Data Scientist & PyCon Speaker
Super Early Bird · first 30 seats
Instant access receipt · Supports Apple Pay, Google Pay & Corporate Cards
✓ 100% Week 1 Guarantee: Run live agents in Session 1 or get a full 1-click refund.