Skip to content

// get started

Start in one command.
Then just talk.

Run one command inside your project, restart your tool, and just talk. MasterMind wires itself into that project (Claude Code, Cursor and Codex) and applies the right discipline automatically.

1

Install, in your project

Run this from inside the project you want MasterMind in. It wires that project for Claude Code, Cursor and Codex, nothing global. It also writes an AGENTS.md, the open instruction file, so any other tool that reads one picks the brain up too. The same command updates it later.

You'll need an AI coding tool first, Claude Code, Cursor or Codex.

bash
$ npx mastermind-brain

before you run it

  • Node 18 or newer
  • Git and Bash
  • On Windows, run it inside WSL. Native Windows is refused, and so is Git Bash, which runs the Windows build of Node.

Want it in every project? add --global. Then restart your AI tool: until you do, the brain isn't loaded. Confirm it's live: ask “are you running as MasterMind?”

2

Set it up, say “init”

Open your project and tell MasterMind to get ready. It sets up once and prints a short “ready” report, which is also your proof it's engaged.

you ▸ init

🧠 MasterMind ▸ detected Next.js · TypeScript · Tailwind

✓ frontend field built for your stack, ready to build

You already have code

It reads your package.json/config, detects the stack, and follows your conventions, nothing to answer.

A blank folder

It asks one open question (“what do you want to build?”) and if you're not technical, it picks the stack for you.

Say it in plain language: “set up MasterMind” or “initialize MasterMind”. Avoid a bare /init in Claude Code, which has a built-in command by that name. Not wired up? Run npx mastermind-brain check.

3

Just talk, no commands to learn

That's the last bit of setup. From here on, say what you want in plain language. MasterMind recognizes the intent, applies the right skill, and shows a one-line note that it did.

you ▸ add a search box to the header

🧠 MasterMind ▸ building this, will verify before handoff

└ build · design · implement · verify · review

✓ debounced, labeled, keyboard-ready, checked before “done”

Skipped step 2? It sets itself up on your first task anyway. Want the full menu? Just ask “what can you do?” (the help skill).

want to check on it?

You never have to run these. MasterMind calls them itself, which is the point: in Cursor and Codex it can now ask for the one skill it needs instead of guessing a file path. They're read-only, so nothing installs or changes when you run them.

.mastermind/bin/mastermind skills : every skill, one line each

.mastermind/bin/mastermind skill performance : the exact instructions it follows

.mastermind/bin/mastermind wrong-log : every time it got something wrong, and what caught it

That last one is the honest answer to “can I trust this?”: a record of misses with the catcher named, never a self-assessment. Empty means nothing was logged, not that nothing went wrong.

working in a monorepo?

One brain lives at the repo root, but each app can get its own field and lessons: the frontend app builds like a frontend, the API like a backend. MasterMind picks the right one by file path, so it never guesses which app you're in.

apps/web → frontendapps/api → backendpackages/ui → frontend · design

// what gets wired

Where it's tested

One npx mastermind-brain wires this project for Claude Code: native skills, agents, and the kernel, loaded automatically. That's where MasterMind is developed and where its measured results come from.

Cursor gets the kernel and the active field pack as always-on rules; Codex reads the project's AGENTS.md. Those three are the supported set: anything else is unwired and unclaimed.

ToolSetup
Claude Codenpx mastermind-brain
Cursornpx mastermind-brain
Codexnpx mastermind-brain

Already installed it globally?

Your setup keeps working, nothing breaks. To switch to per-project: run npx mastermind-brain uninstall --global, then npx mastermind-brain inside each project you want.