Hi, it's Ivo. Most videos about REA point it at someone else's app. I'd start with my own business. Your website sends its code to every visitor's browser, so whatever is in that code, a stranger can read too. REA just makes it fast to see.
Everything below follows REA's own docs.
The one rule first
Only your own website, or one you have written permission to check. That's the project's own rule: REA is for "software you own or are authorized to examine" (README and DISCLAIMER.md on github.com/morluto/rea).
Their disclaimer also says it plainly: "Public availability of software ... does not by itself establish permission." A site being online doesn't mean you may examine it. If an agency built your site and hosts it, ask them before you start. This is not legal advice. When in doubt, don't.
What you need
- Claude Code (it needs a paid Claude plan: Pro, Max, Team or Enterprise).
- Node.js: REA asks for version 22.23.4 or newer in the 22 line (or 24.22.1+, or 26.12.1+). Type
node -vin your terminal to check. - Chrome (or another Chromium browser like Brave). REA's website checks attach to a separate Chrome window.
- REA itself. In your terminal run
npx rea-agents setup, pick Claude Code, read the plan, approve it, and restart Claude Code.
The own-website check prompt
Open Claude Code in an empty folder and paste this. Put your own address where it says so.
Use REA to check my own website: [paste your website address]. I own this site. Start a separate Chrome window with a fresh test profile for this, on a local debugging port only, and close it when you're done. Don't log in, don't fill in forms, and don't send anything. In plain English, list: 1. The page addresses you can find in the site's code, including any that aren't in the menu. 2. The outside tools and services the site loads. 3. The servers and web addresses its code talks to. 4. Whether the site publishes source maps that show the original code. For each item, show the evidence and say how sure you are. Mark anything you're guessing.
Claude will ask before it starts the browser and runs REA's tools. Read each step before you say yes.
What each finding means
- Page addresses in the code. REA's docs call these "route candidates". An old offer page or a test page that's no longer in the menu can still be listed here. Not every address is a real, open page.
- Outside tools. The docs call these "vendor fingerprints": chat widgets, analytics, payment or booking tools your site loads. It's a good moment to ask if you still use all of them.
- Servers it talks to. "Endpoint candidates": the web addresses your site's code calls. Anyone can see these. What matters is that they only answer what they should.
- Source maps. These files map the shrunk code a browser gets back to the original code. MDN describes them as "allowing the original code to be reconstructed". If your site publishes them, anyone can read your original code. Many builders turn them off for the live site.
The message to send whoever built your site
Copy, fill in, send:
Hi [name], I ran a check on our own website with an AI tool called REA. It listed a few things I'd like you to look at: - Page addresses in the code: [paste] - Outside tools the site loads: [paste] - Servers the site talks to: [paste] - Source maps: [found / not found] Can you tell me which of these should be public, and remove anything that shouldn't? Thanks.
The honest part
- REA's own disclaimer says results "may be incomplete or incorrect". Treat the list as questions for your builder, not as proof that something is wrong.
- REA runs on your computer, but Claude reads the results. Your Claude plan's data rules apply.
- Use what you find to fix your own site. Don't use it to copy anyone else's.
- REA changes fast. If something fails, run
npx rea-agents@latest setupagain.
As of October 26, 2026, REA has over 380,000 stars on GitHub. Menus, versions and plans change.
Sources: github.com/morluto/rea (README, DISCLAIMER.md, docs/browser-observation.md), read October 26, 2026; developer.mozilla.org (Glossary: Source map).