// T3 · Design and run systems of agents
AI Orchestrator Training
One assistant doing one job gets you far. Past that, the hard part stops being the prompt and becomes the system: a spec it is built from, which agent does what, which tools each one may call, what fits in its context, where a person checks, how injected instructions are kept from doing harm, and what it all costs. This training teaches you to design those systems, run them in production and hand them over. It covers skills and MCP (the standard way to connect an agent to other systems) in plain terms. The Implementer and Practitioner trainings are included, so the whole path is there if you need it.
For Implementers, technical consultants and in-house leads who build or run AI across several steps, tools or teams
// What you walk away with
Write a spec an agent system is built and tested from, and choose between one agent, a workflow and several agents
Design tools, skills and connections so agents call the right tool with the right inputs, without flooding their context
Guard agents against prompt injection and overreach, with checks in code and a person where the risk is
Test, trace and monitor the system, and keep every run inside a budget
Run it for a client or your own company, within EU rules, with an owner, a runbook and safe changes
// Inside the training
From spec to system
Spec Driven Development, when one agent is enough and when to split the work, and the orchestration patterns and handoffs that tie several agents together.
Tools
Designing tools an agent uses correctly, how many to expose, controlling chained tool calls, and when to use skills or MCP.
Context and memory
Why agents get worse as their context fills up and how to prevent it, what to remember across runs, and agents that run for hours or wait on people.
Guardrails and security
Guardrails at every layer, prompt injection and the three conditions that make it dangerous, least access, and vetting outside servers and skills before you connect them.
Run it
Evals built from the spec, a trace of every run, the few signals worth watching, and keeping cost and waiting time inside a budget.
Change and hand over
Safe, reversible changes, what the EU AI Act and GDPR mean in practice, and a runbook and owner so the system keeps working after you step away.
// Step 2 · Live class, optional
Apply it live, with a trainer.
The online course teaches the method. The live class is where you use it: two and a half hours online, in a small group, working on a task from your own week with a trainer on hand. Pick the date that suits you.
It costs €495 added together with the course, or within 24 hours of buying it. Added later, it is €645. Buying the course now? Pick a date in the box above. Already have the course? Book a seat here.
The next class dates are being scheduled.
Leave your email and hear about new dates first.
// This is for you if
You already have AI doing single jobs well, and now need several agents and tools to work together reliably and safely, for clients or for your own company.
// Look elsewhere if
You want a first quick win in your own work. The Practitioner training gets you there faster and costs less.
Built from agent systems running in production, not theory. The Implementer and Practitioner trainings are included, so you can start from the beginning.
// Design and run systems of agents
€995 one-time · instant access
// Prefer it built for you?
Skip the learning curve.
If you would rather hand the whole thing over, the done-for-you packages and the AI Deep Scan are one page away.
Not ready to talk? Take the 2-minute readiness scan →