Skip to content

Get started

The five-minute setup. Install the tool, add the team's plugins, then wire the repo so every agent — yours and your teammates' — starts already knowing the project. Each step links to the deep dive; this page is just the checklist.

The whole setup, start to finish — 1 min 48 s.

For you · your machine

Set up your machine

  • 1 · Install Claude Code

    curl -fsSL https://claude.ai/install.sh | bash
    Then cd your-project && claude and sign in once.

    Full setup & first session

  • 2 · Add the core plugins

    /plugin install superpowers@claude-plugins-official
    Then code-review, frontend-design, ralph-loop, context7. /reload-plugins.

    Full catalog

  • 3 · Pick the model

    Default to Opus 4.8 with high thinking for good quality.

    Choosing models

For your team · the repo

Set up the repo

One source of truth, loaded automatically at the start of every session. Keep a thin root that just points at AGENTS.md, and write the real content there once.

CLAUDE.md      →  @AGENTS.md          (Claude Code reads this)
.cursorrules   →  see AGENTS.md       (Cursor reads this)
AGENTS.md      →  the single source of truth

Big repo? Keep it lean. Have AGENTS.md hold a short overview + shared commands, then point to focused files loaded only when the task needs them — FRONTEND.md, BACKEND.md, DESIGN.md. One giant memory file taxes every session.

Put in AGENTS.md: the exact test / build / lint commands, code & comment standards, a short module map, architecture boundaries, and known failure modes.

Memory & project rules — the full pattern

If you remember one thing

Abuse markdown files

Write everything down in .md files — plans, rules, decisions, scratch notes. Durable memory the model reloads every session is the best way to work with an LLM, and the direct fix for its biggest weakness: it forgets.

Checklist

  • Claude Code installed and signed in
  • Core plugins installed: superpowers · code-review · frontend-design · ralph-loop · context7
  • Model set to Opus 4.8, high thinking
  • AGENTS.md created with rules, standards, and test/build commands
  • CLAUDE.md and .cursorrules point at AGENTS.md
  • (Big repo) AGENTS.md links out to FRONTEND.md / BACKEND.md / DESIGN.md