🦾 Harness — agent's armor
The model is just a brain. Tools, system prompts, memory, and safety rules are what allow it to do things in the real world.
Agent Builder · Learning game
A brain in a jar.
Build an agent.
Three chapters, three instructive failures, and nine simulation missions: harness a raw language model (Harness), adjust its execution loop (Loop), and manage its memory budget (Context). Your progress is saved right here in your browser.
Start playing
Why try this product?
A three season game where each season starts with a failure.
Instead of defining, you first see what breaks without Harness, Loop and Context; Then you make it yourself. Earn XP, unlock each chapter's badge, and receive a "Complete Agent Builder" certificate at the end.
- 9 simulation missionsFrom simple to advanced; See the result of each decision in the execution report.
- XP, badges and certificatesYour progress will be saved and you can continue from there.
- About 45 minutesNo coding; For someone new to agents.
Full access, free
Do you want to learn more and differently?
First, register your profile and email address here to get full access to this educational content, simulators, and games — once for all builds.
No spam; Only the news of new creations and courses. Do you have an account? enter
the concept
Learn with your hands.
Most tutorials give definitions first, then examples. The agent works the opposite way: every season with The story of a failure You start to see what would go wrong without that concept; After a short lesson, then three missions in which you create your own agent and see the consequences of your decisions in the execution report.
With 4 correct answers out of 6 test questions of each chapter, you get the badge of that chapter and the next chapter opens. At the end, a combined mission and "Perfect Agent Maker" certificate await you.
🔁 Loop — execution loop
Decision, implementation, observation, review. A good loop knows when to try again and when to stop; in Loop engineering workshop Practice it more deeply.
🧠 Context — memory budget
Everything the model sees passes through a limited window. Selection, summarization and timely elimination, ibid Context engineering is
Where is my progress saved?
Only in the local memory of this browser (localStorage). No data is sent to the server; If you clear the browser cache or log in with another device, you will start from the beginning. The "reset" button on the top bar of the game does the same thing.
Who is this game for?
For someone who is new to artificial intelligence agents and wants to have the right mindset of three basic concepts before building the first agent. If you work with tools like Claude Code, n8n or LangChain, compare the missions with your everyday experience.
From game to project