- Home
- Build AI agents without code
AI agent training
Teach your teams to build AI agents without code
In Build an Agent, employees learn to build AI agents without code: a role, rules, tools and skills written in plain words, then tested on real customer messages. AI agent training by doing.
- No code
- Tested on customer messages
- Final assessment
What an AI agent is, in plain words
An agent is an assistant with a role, rules it must follow, tools it can use and skills it has been taught. Learners write all of it in plain words, and talk to their agent as a customer would.
- A role. Who the agent is and who it helps.
- Rules and limits. What it does, what it never does, and what it does when information is missing.
- Tools and skills. What it can look up or do, and what it has been taught.
- 1Give it a role
- 2Set its rules
- 3Add a tool
- 4Teach it a skill
- 5Test it on customers
Short sessions to build AI agents without code
The program has 13 short sessions, two prompt games and a final assessment, about 45 minutes each.
Meet an AI Agent, then give it a role
What an agent is, and how a role changes what it does.
Set its rules and limits
Boundaries, a voice, and what to do when information is missing.
Teach it facts, tools and skills
Give it its first tool, a second one to choose from, then a skill.
An agent that acts
Confirm first: an agent that does things stays under the learner's control.
Tested on real customer messages
A learner tests their agent in a playground with messages a customer would send, and changes its instructions until it answers well.
Defended against prompt injection
Anyone building agents must know how they get tricked.
Crack the Vault
Get a password out of an assistant that guards it better at every level.
Lock It Down
Attackers want a secret from the learner's agent: make it hold, and keep helping real customers. Scored when the attack ends.
Don't let your agent get tricked
Rules that hold under pressure, written by the learner.
A final assessment: their agent, live
A free project
Scored on what the agent does when the learner hands it in.
Nothing to install
Everything runs in the browser, on a real agent, in a sandbox of their own.
No admin action
Assessment is scored by the platform; nobody has to unblock a learner.
Questions about building AI agents without code
Do learners write any code?
No. An agent's role, rules, tools and skills are written in plain words, and learners never see code on their screen.
How is the agent tested?
In a playground, on messages a real customer would send. A milestone is a test of the agent, or a change to its instructions.
Does the training cover security?
Yes: Crack the Vault, Lock It Down and Don't Let Your Agent Get Tricked train defending an agent against attempts to extract a secret or bend its rules.
Which agent do learners build with?
Sessions are built with a real agent run by the platform: Claude Code, Codex or GitHub Copilot, as your organization chooses.
See agent training with your own teams.
A 30-minute call: we get to know your project, then show you the platform.
Book a demo