Hands-on, task-focused guides for someone who's been given a Brain URL and a token. Each tutorial is self-contained, takes ≤15 minutes, and ends with a working capability.
If you're installing External Brain yourself (operator role), read
../QUICKSTART.md and ../USING_BRAIN.md
first — they cover the server side.
| You want to… | Tutorial | Time |
|---|---|---|
| Wire your AI tool (Claude Code / Cursor / Windsurf) to your Brain for the first time | 01 — Getting started | 10 min |
| Ask your Brain questions about your own coding history | 02 — Asking the Oracle | 10 min |
| Teach the Brain a new pattern, rule, or preference | 03 — Teaching the Brain | 10 min |
| Issue a token scoped to a specific organization or project | 04 — Token scope + management | 10 min |
Export your accumulated rules into a project's .claude/ / .cursor/ / AGENTS.md |
05 — Exporting rules | 5 min |
| Diagnose "the Brain doesn't seem to be helping" | 06 — Troubleshooting | as needed |
| Understand the skill types (Recipe, Rule of thumb, Reflex…) — no tech background needed | 07 — Skill types, explained | 10 min |
Your AI coding tool talks to your Brain over MCP-over-HTTP. Before the model generates code, the Brain quietly injects relevant rules from your past sessions ("we use react-hook-form here, not Formik"). After the session, the Brain extracts new rules from what worked. Over time the loop tightens — your tool stops repeating the same mistakes, you stop re-explaining the same constraints. The webapp is where you see what the Brain knows, ask it questions, and curate the rules.
The deeper "why" lives in ../HOW_IT_WORKS.md;
these tutorials are the "how".