AI agents
Describe the lab. Agents build it.
A 3-hour Kubernetes lab for 40 learners on Friday, scored.
- Lab Builder Plan ready: 3 milestones, 3 hours, 40 seats
- Image architect Golden image: Ubuntu 24.04, kubectl, kind, Helm
- Track Enricher 18 steps written, grounded in our MCP knowledge
- Scoring Engine Scorecard: 9 checks across 3 skill areas
- Inventory 40 lab accounts reserved for Friday 09:00
The agent mesh
Nine agents. One team.
Each agent has one job and hands its work to the next. Together they take a lab from idea to the live room.
Lab Builder
Turns "a 3-hour Kubernetes lab for 40 people on Friday" into a full lab plan you can build with one click.
Track Enricher
Writes the overview, briefing, milestones and every step, grounded in our proprietary MCP knowledge and checked by a quality gate.
Track Author
Starts a brand new track from your outline, your guide or a document you upload.
Image Architect
Designs the golden image: operating system, tools, packages and data, ready to build.
Scoring Engine
Drafts a scorecard with skill areas, weights and a check for every task.
Dataset Builder
Generates realistic sample data so every lab has something real to work on.
Event Coordinator
Creates the event, connects tracks and lines up seats for the day.
Proctor Agent
Watches every live lab, spots who is stuck or idle, and raises an alert for a human.
Mentor Agent
Guides a stuck learner with hints and questions, without handing over the answer.
How it works
You describe. They draft. You decide.
- 01
Describe
Say what you want in plain words, or upload an existing guide.
- 02
Draft
Agents plan the lab, write the steps and draft the checks.
- 03
Review
Edit anything in Track Studio. Your edits always win.
- 04
Deliver
Publish, and the lab is ready for one person or a whole event.
Track Enricher
Complete lab guides, written for you.
A short brief becomes a full track: overview, briefing, milestones and step-by-step instructions with sources. A quality gate checks every milestone and rewrites the weak ones.
- Grounded in our proprietary MCP knowledge
- Bring your own guide as a starting point
- Re-enrich later and keep every human edit
Package the web app as a container image so it runs the same everywhere.
$ docker build -t web:1 . Keep the image small: start from an alpine base.
Image web:1 exists on the lab machine
Proctor and Mentor agents
An extra pair of eyes in every lab.
The Proctor agent watches every session and tells you who needs help. The Mentor agent talks the learner through it with hints, not answers. You can always step in with Shadow or Assist.
Live proctoringapiVersion: apps/v1 kind: Deployment spec: replicas: 3 template: spec: containers: - image: web:1 ports: [8080]
AI helps. It never grades.
Scores come from checks on the real lab environment, never from a model’s opinion. Agents draft content and help people, and a human publishes every lab.
FAQ
Questions people ask
Do I have to use the agents?
No. You can write every lab by hand in Track Studio. The agents are there to save time, and you can switch them on or off per organisation.
Where does the content come from?
The agents research and write using our proprietary MCP knowledge layer, plus any guide or document you upload. Every step can be edited before you publish.
Will re-running the agents overwrite my changes?
No. Re-enrich keeps your edits and only fills in or improves the parts you have not touched.
Does AI decide a learner’s score?
Never. Scores come from automated checks on the real environment. AI is not in the scoring path.