Installation

3 ways to install

Choose the path that fits your workflow:

1. CLI (developers with a terminal)

npm install -g lytos-cli
lytos init

The CLI detects your stack, asks your project name and AI tool, then scaffolds everything. Interactive or with flags:

lytos init --name "My API" --tool claude --yes

2. Install script (lightweight, no npm)

curl -fsSL https://raw.githubusercontent.com/lytos/lytos/main/install.sh -o install.sh
bash install.sh

Same result as the CLI, but downloads files from GitHub instead of using the npm package.

3. GitHub template (vibe coders, no terminal)

Use the starter template → click “Use this template” → open in Claude Code or Codex → say “Help me configure Lytos.”

Perfect for developers who code directly from AI web apps.

What gets created

.lytos/
├── LYTOS.md              ← AI briefing (read once)
├── manifest.md           ← your project's constitution
├── memory/
│   ├── MEMORY.md         ← memory index
│   └── cortex/           ← specialized zones
├── skills/               ← 9 operational procedures
├── rules/                ← quality criteria
├── issue-board/          ← Kanban board (folder = status)
├── templates/            ← sprint template
└── scripts/              ← automation tools

Plus a CLAUDE.md or .cursorrules at your project root (depending on your AI tool choice).

After installation

Open your AI tool and say:

“Help me configure Lytos for this project.”

The AI reads the briefing, understands the method, and guides you through filling the manifest.