A beginner-friendly guide to installing Leopoldo on Claude Cowork. From downloading the package to your first conversation with the orchestrator, every step explained.
Leopoldo Editorial
This guide walks you through installing Leopoldo on Claude Cowork from scratch. No technical background required. Every step is explained.
Claude Cowork is Anthropic's collaborative AI workspace. Unlike Claude Code (which runs in a terminal), Cowork has a visual interface where you interact with Claude through a chat-like experience. It is designed for professionals who want AI assistance without touching the command line.
Leopoldo is an expertise system that makes Claude a specialist. Out of the box, Claude is a generalist. Leopoldo adds deep domain knowledge (finance, legal, consulting, development) plus an orchestrator that routes every request to the right expert automatically.
You choose a domain. You get the full system.
Make sure you have:
That is everything you need.
After requesting access at leopoldo.ai, you receive an email with a download link. Click it to download the ZIP file. The link is valid for 7 days and allows up to 3 downloads.
Make sure you received the Cowork version (not the Claude Code version). If you are not sure, check the email or reply to it asking for the Cowork format.
Open the ZIP file and extract (unzip) its contents directly into your workspace folder. The result should look like this:
The key thing: the .claude-plugin folder must be at the root of your workspace. Not inside a subfolder.
**On macOS:** Double-click the ZIP, then drag the extracted contents into your workspace folder.
**On Windows:** Right-click the ZIP, select "Extract All", then copy the extracted contents into your workspace folder.
If your workspace already has a .claude-plugin folder, that is fine. The installer merges everything safely.
Launch Claude Cowork and open the workspace folder where you extracted the files.
On the first session, Leopoldo detects the extracted files and installs itself. This happens automatically. You will see the orchestrator set up hooks, create the manifest, and load your domain expertise.
This only happens once. On future sessions, everything is already in place and Leopoldo activates instantly.
Now try a simple test. Type something related to your domain in the chat:
If the orchestrator routes your request to a specialized agent and you get structured, professional output, everything is working.
Cowork supports slash commands. Type these in the chat:
Cowork also generates domain-specific commands automatically. Depending on your domain, you might see commands like /due-diligence, /valuation, /compliance, or /client-report. These are shortcuts that invoke specific capabilities directly.
When you open a workspace with Leopoldo installed in Cowork:
1. Leopoldo detects the extracted files and runs the installer (first session only) 2. The orchestrator activates and reads your domain expertise 3. Every message you send is classified by intent 4. The orchestrator routes to the right specialized agent 5. Quality gates verify the output before you see it
You do not need to think about any of this. Just chat with Claude normally.
The capabilities are identical. The same domains, the same orchestrator logic, the same quality gates. The differences are only in the interface:
Choose whichever you are more comfortable with. Many users start with Cowork because it feels more familiar.
**Claude does not seem to know about Leopoldo.** Make sure the .claude-plugin folder is at the root of your workspace. Try closing and reopening the workspace in Cowork.
**I accidentally installed the Claude Code version.** Delete the .claude folder and extract the Cowork version instead. Reply to your welcome email if you need the correct format.
**Slash commands are not showing up.** Close and reopen Cowork. The commands are generated when Cowork first loads the plugin.
**I want to start fresh.** Type /leopoldo uninstall in the chat, then extract the package again.
**I have questions.** Email hello@leopoldo.ai or book a call at cal.eu/leopoldo.ai/discovery-call.
You are ready. Start chatting with Claude in your domain. The orchestrator handles everything: routing, quality checks, and session management.
Explore all domains at leopoldo.ai. Full Stack is free on GitHub. Specialized domains available on request at hello@leopoldo.ai.
You need Claude Cowork installed and a workspace where you want to use Leopoldo. Cowork requires a Claude subscription from Anthropic.
Claude Code runs in your terminal (command line). Cowork runs as a visual application with a graphical interface. Leopoldo works on both, but the installation folder structure is slightly different. Choose the guide that matches your platform.
No. Cowork is designed as a visual interface. You talk to Claude in plain language, and the orchestrator routes your request to the right specialist. No terminal, no commands, no programming.
No. The folder structure is different between platforms. When you request access, specify which platform you use and you will receive the correct package. If you use both, you can request both formats.
Type /leopoldo update in the Cowork chat. Updates are always explicit and never happen automatically. Your customizations are preserved.
Yes. Each domain adds its own expertise. The orchestrator knows which specialist to use based on your message. Use /leopoldo add to install additional domains.