Setup Guide: Install Leopoldo on Claude Code (Step by Step)

This guide walks you through installing Leopoldo on Claude Code in 5 minutes. No technical background required.

Install in one command (recommended)

After Leopoldo sends you the welcome email, click Get started. You land on a branded page that shows a single command. Copy it, open Terminal, change into your project folder, and paste it.

curl -fsSL \
  "https://leopoldo-api-production.up.railway.app/api/licenses/install-script?token=YOUR_TOKEN&flavor=claude-code" \
  | bash

The installer:

1. Asks for confirmation before changing anything. 2. Refuses to overwrite an existing .claude/ or .leopoldo/ unless you set LEOPOLDO_OVERWRITE=1. 3. Downloads your tailored bundle (counts as one of your 3 ZIP uses). 4. Extracts skills, agents, hooks, and brand kit into the current folder. 5. Prints the next steps: run claude, then /leo-onboard.

The script is plain shell with no hidden behaviour. The welcome landing page also shows the same command so you can copy it directly from your browser.

Install manually (alternative)

Prefer to inspect the bundle before extracting, or unable to use the one-command installer? Follow the steps below. The end result is identical to the one-command install.

What Is Claude Code?

Claude Code is Anthropic's official command-line tool for working with Claude. It runs in your terminal and gives Claude access to your files, your project structure, and your local tools. If you do not have Claude Code yet, install it first from docs.anthropic.com (Claude Pro, Team, or Enterprise subscription required).

What you have

When you purchase one or more Leopoldo domains, the download link in your welcome email delivers a single ZIP containing:

• Leopoldo Core (always included). Identity contract, orchestrator, brand profile, attribution, onboarding, learning layer.
• Your domain plugins under .claude/plugins/. One folder per purchased domain (Investment, Advisory, Legal, Consulting, etc.).
• `leopoldo-client.json`. Your personalized configuration: activation key, entitled plugins, expiry. Stays inside .claude/ and powers /leopoldo update.
• CLAUDE.md with the Leopoldo section between markers (merged into your existing CLAUDE.md, if any).

One extraction installs everything at once. No per-plugin upload step on Claude Code.

Step 1: Install your plugins (one-time, 3 minutes)

1a. Download and extract the archive

Click the Download your system button in your welcome email. You get a single ZIP. Extract it directly into your project folder. The result looks like:

your-project/
├─ .claude/
│  ├─ plugins/
│  ├─ skills/
│  ├─ agents/
│  ├─ commands/
│  └─ leopoldo-client.json
├─ .leopoldo/
└─ CLAUDE.md

The .claude/ folder MUST be at the root of your project (alongside .git/, package.json, or whatever else lives there). Not in a subfolder.

macOS: double-click the ZIP, drag the extracted contents into your project folder. Windows: right-click → Extract All → copy contents into your project folder. Linux: unzip leopoldo-*.zip -d /path/to/your-project in your terminal.

If your project already has a .claude/ folder, that is fine. Leopoldo merges safely between markers in CLAUDE.md.

1b. Activate on first launch

Open your terminal, navigate to your project folder, and start Claude Code:

cd /path/to/your-project
claude

On the first session Claude Code asks you to confirm activation. The activation key from leopoldo-client.json is read automatically, so the prompt is usually just Enter to confirm. If for any reason the file is missing, paste the key from your welcome email when asked.

This activation only happens once. On subsequent sessions Leopoldo loads instantly.

Step 2: Start onboarding

Inside Claude Code, type:

/leo-onboard

Or say "Ciao Leo vorrei fare l'onboarding" in plain language. Leopoldo introduces itself, confirms the install, and walks you through a 5-minute setup:

• Confirm your brand profile (defaults to the Leopoldo signature; switch to a client profile if needed)
• Pick your primary language (mirrored across all deliverables)
• Produce your first sample deliverable so you see the full pipeline in action

After onboarding, just talk to Claude in your domain language. The orchestrator routes each message to the right specialist automatically.

Test it

Try a domain-specific prompt to verify the install:

• Investment Core: "Build a quick DCF for Apple Inc"
• Legal Suite: "Draft an NDA for a software consulting engagement"
• Senior Consultant: "Size the European market for AI-powered legal tools"
• Engineering: "Set up a Next.js app with authentication and tests"

You should see the orchestrator route to a specialized agent, run quality gates, and return a structured deliverable. If you do not, jump to Troubleshooting below.

Key commands

Leopoldo adds these management slash commands inside Claude Code:

• /leo-onboard: start (or restart) the onboarding
• /leo-brand: switch brand profile, list active profile, create a client white-label
• /leo-introspect: ask Leopoldo what it just did in this session (ledger + skills + agents)
• /leopoldo status: check installed plugins and integrity
• /leopoldo update: pull the latest version (explicit, never automatic)
• /leopoldo repair: reinstall missing or corrupted files
• /leopoldo add <domain>: add another domain alongside your current install
• /leopoldo uninstall: remove everything cleanly, restore previous CLAUDE.md from backup

Note: /leo-* are the new short-form slashes (since v1.15). The longer /leopoldo * forms remain valid for the management commands (status, update, etc.).

How it works behind the scenes

On every Claude Code session start in a Leopoldo project:

1. The session-start hook reads .claude/manifest.json to identify managed vs user-owned files. 2. An offline integrity check verifies every managed file is intact. 3. The orchestrator loads identity (canonical from leopoldo-router), brand profile from .brand/active.txt, and the domain agents available. 4. Your prompt is classified by the router and dispatched to the right workflow agent. 5. Quality gates (citation, attribution, completeness) run before output reaches you. 6. Output is rendered with the active brand: Warm Black + Ivory + Hunter Green + Instrument Serif if you stayed on the Leopoldo default, or fully white-label if you switched to a client profile.

You do not need to think about any of this. Just talk to Claude normally.

Updates

Updates are explicit, never automatic.

• /leopoldo update checks for new versions via the Leopoldo backend, creates a rollback snapshot, applies new managed files, leaves your customizations untouched.
• After update, the orchestrator confirms what changed (new skills, agent retrofits, brand presets) in a one-paragraph summary.
• If anything misbehaves: /leopoldo rollback restores the previous snapshot in seconds.

Troubleshooting

• Claude does not seem to know about Leopoldo: make sure the .claude/ folder is at the root of your project (run ls -la in your terminal). If it is, type /leopoldo status to surface any missing-file warning.
• Identity is wrong (Leopoldo answers as "Claude"): this means leopoldo-router did not load. Run /leopoldo repair. If the problem persists, reply to your welcome email.
• /leo-onboard does not appear in autocomplete: restart Claude Code (close terminal session, reopen). Slash commands are scanned on session start.
• Activation prompt loops or fails: the activation key in leopoldo-client.json may be stale (e.g., after a renewal). Paste the latest key from your welcome email manually.
• I want to start fresh: run /leopoldo uninstall, then re-extract the ZIP into your project. Your CLAUDE.md is restored from backup.
• Something else is broken: email hello@leopoldo.ai or book a call at cal.eu/leopoldo.ai/discovery-call. We respond within one business day and can rollback in minutes.

Next steps

You are ready. After onboarding, every conversation routes through the orchestrator to the right specialist. The brand profile keeps every deliverable on-style, and quality gates run before output reaches you.

Explore the full catalog at leopoldo.ai. Full Stack is free on GitHub. Specialized domains are available on request at hello@leopoldo.ai. Adding a new domain regenerates your archive on the next download, so a single /leopoldo update brings the new plugin into your existing install.