AI Onboarding Kit

Set up the tools, learn the concepts, and build one thing well — a kit you run once and pass on to your team.

This kit gets one lead at Edventure comfortable with three AI platforms and the ideas behind them, then that lead trains colleagues and interns from the same materials. That’s the whole model: train the trainer.

The three tracks

Three independent platforms you can explore in any order. They reinforce each other but install differently.

TRACK 1

Hermes
A self-improving agent with persistent memory, in a browser dashboard. Runs on Ollama Cloud — no GPU needed. → start here

TRACK 2

Claude Code
Claude in browser VS Code — the everyday driver for spec-driven building. → runbook

TRACK 3

The LaTeX project
A working starter app you extend yourself: a maths prompt → a typeset image. → the build

Your running services

You’re reading this because the kit is already up. These are live on your machine right now — open any of them:

TRACK 1

Hermes dashboard
localhost:9119 · log in trainee/hermes
Skills, profiles, memory, tools, cron.

TRACK 2 & 3

VS Code + Claude Code
localhost:8080
Browser IDE; latex-starter + hermes-home open.

TRACK 3

LaTeX app
localhost:3584
The running starter (after npm run dev).

LESSONS

This onboarding site
localhost:4321
Where you are now.

Starting, stopping, or re-running the kit (./bootstrap.sh) is covered in the repo README — the thing you read before this site was running.

What success looks like

You can install all three platforms, explain skills / profiles / memory / spec-driven development in your own words, run the LaTeX starter and see an image, and walk a colleague through the same path.