Skip to main content

Quick Start

This guide shows how to get started with Rabetbase CLI 2.0 in both new and existing projects.

The current content has been verified against rabetbase-cli v2.2.2 and @lovrabet/cli-framework v1.0.3. Recent versions bundle a jq binary directly in the CLI, so --jq resolves in the order JQ_PATH -> bundled jq -> jq on PATH; write commands keep their preview and confirmation flow, which works well for Agent-driven development.

If you are already using AI coding tools such as Claude Code, Codex, Cursor, Trae, or Windsurf, you can simply copy the prompt below into your AI assistant. It will install the CLI, install the Skill, and handle authentication — you only need to help with browser authorization or project confirmation when prompted.

Prompt (copy and paste into your AI assistant):

Please install Rabetbase CLI: https://www.lovrabet.com/agent-install-guides/rabetbase-cli.md

The most common point where Agent-based setup pauses is browser authorization: when the Agent prints an actual Login URL, open it, complete authorization in the browser, then return and say "authorization completed."

If the Agent asks whether to initialize a project, confirm only when the current directory is the real target project. Installing the CLI alone does not require rabetbase project init.

Supported AI Agent tools: Claude Code, Codex, Cursor, Trae, Windsurf, GitHub Copilot, OpenCode, Hermes Agent, OpenClaw

💡 If you prefer to control each step yourself, or do not use an AI Agent tool, continue with the manual installation below.

Manual Install​

For: First-time Rabetbase CLI users on a machine that has never installed any previous version.

>> Click here for the installation video walkthrough

1.1 Install the CLI​

Package: @lovrabet/rabetbase-cli

# Global install (recommended)
npm install -g @lovrabet/rabetbase-cli@latest

# Verify the installation
rabetbase --version

Latest version: npm version

1.2 Install the Rabetbase Skill​

The recommended way is to use the built-in CLI entry:

rabetbase skill install

If you prefer the raw skills command, you can still run:

npx skills add lovrabet/rabetbase -g -y

A global install is recommended so AI assistants can use rabetbase from any project directory.

1.3 Authentication​

rabetbase auth login

A browser window will open for OAuth login. The CLI saves the session automatically, and one login works across projects.

1.4 Initialize a Project​

Initialize an existing project directory:

cd your-project/
rabetbase project init

Create a brand-new project from scratch:

rabetbase project create

1.5 Quick Verification​

rabetbase doctor
rabetbase schema --format compress
rabetbase app list --format compress

If you want to verify that --jq is already available, run:

rabetbase schema --format compress --jq '.data.cli.version'

2. Upgrade from CLI 1.x​

For: Machines that still have Lovrabet CLI 1.x installed, with older projects using the legacy CLI.

2.1 Install the New CLI​

# Global install (recommended)
npm install -g @lovrabet/rabetbase-cli@latest

# Verify the installation
rabetbase --version

2.2 Remove the Old Lovrabet CLI 1.x​

npm uninstall -g @lovrabet/cli

2.3 Auto-upgrade Old Projects​

>> Click here for the upgrade screencast demo

Run this in each old project directory:

rabetbase project upgrade

The command automatically performs the following migration steps:

StepAction
1.lovrabet.json -> .rabetbase.json
2Delete .lovrabet/
3Remove legacy IDE Skill files
4Clean up old MCP config (lovrabet-dataset)
5Remove the old Skill lovrabet/lovrabet-skill
6Install the new Skill lovrabet/rabetbase

2.4 Manual Fallback Upgrade​

If the auto-upgrade fails, use this sequence:

# 1. Remove the old Skill
npx skills remove lovrabet

# 2. Install the new Skill
rabetbase skill install

If your Claude Code or IDE still contains the old MCP entry, remove it as well:

claude mcp remove lovrabet-dataset

3. Most Important Changes in the Current Version​

  • --jq is now more reliable: The latest build bundles jq as a sidecar, so installation no longer depends on node-jq downloading binaries from GitHub.
  • Dataset lists are more complete by default: rabetbase dataset list returns DO V2 datasets, including both DB_TABLE and METADATA. Use --source DB_TABLE for database-backed datasets, or --source METADATA for metadata datasets.
  • AppCode environment variables must be passed explicitly: CI can still store RABETBASE_APPCODE, but commands should use --appcode "$RABETBASE_APPCODE". The CLI no longer lets an environment variable silently override project config.
  • Writes are safer by default: rabetbase page generate-start now returns a dry-run preview by default, and a real generation job requires --apply; project-scope writes also avoid leaking global app config into project files.
  • Dataset write commands must target explicitly: dataset rename, field-update, extend-update, and delete require --code, --id, or --dbid, and should be previewed with --dry-run first.
  • Menu resource updates are more conservative: rabetbase menu update preserves existing page extend config while updating CDN resource URLs, reducing the risk of overwriting production settings.
  • You can inspect real data directly from the dev CLI: rabetbase data filter and rabetbase data getOne are now part of the official command surface.
  • Skill installation is more consistent: You can now simply run rabetbase skill install without remembering the lower-level npx skills add ... form.

4. Checklist​

Fresh Install​

  • Install CLI: npm install -g @lovrabet/rabetbase-cli@latest
  • Install Skill: rabetbase skill install
  • Authenticate: rabetbase auth login
  • Initialize the project: rabetbase project init or rabetbase project create
  • Verify: rabetbase doctor

Upgrade from 1.x​

Per machine (do once per machine)

  • Install the new CLI: npm install -g @lovrabet/rabetbase-cli@latest
  • Remove the old CLI: npm uninstall -g @lovrabet/cli
  • Remove the old Skill: npx skills remove lovrabet
  • Remove the old MCP: claude mcp remove lovrabet-dataset
  • Install the new Skill: rabetbase skill install
  • Re-authenticate: rabetbase auth login

Per project (do once per project)

  • Auto-upgrade: rabetbase project upgrade

5. Next Steps​