CONNECT YOUR AGENT
Rat Radar is an MCP server: add it once and your AI app can check any site, get a playbook and X-ray a page by itself. Free, no key, read-only. Pick your app:
Rat Radar tells a person or an AI agent whether a website will let them collect its data, and the best way to do it.
MOST HOSTILE NOW
x1.0
SCROLL OR + - TO ZOOM / DRAG TO PAN / CLICK A BUBBLE, COUNTRY OR DOT
SITEVERDICTHOSTILITYWHAT TO DO
SIGNAL
EXTRACTION PLAN
PASTE ANY URL ABOVE
X-Ray reads the page the way a scraper would and tells you the best way to pull its data.
- Ranked sources: JSON-LD, hidden JSON, repeated cards, tables, feeds
- The fields and selectors for each, with real rows extracted as proof
- Pagination, blockers, and starter code in Python, Playwright or Apify
Tip: from the Radar tab, open a site and press X-RAY THIS SITE.
Rat Radar tells a person or an AI agent whether a website will let them collect its data, and the best way to do it.
AGENT TOOLS AND LIVE USAGE
What an AI agent can call once it is connected, with an example of what to ask it. Counts are live from the MCP server: totals only, nothing personal.
WHO IT IS FOR
- Founders and sales teams who want leads, prices and signals from public websites.
- Analysts and researchers who need a clean dataset and want to know the cheapest legal route to it.
- AI agents that must decide, before every fetch, whether and how to collect from a site.
HOW TO USE IT TODAY
- Open RADAR, find a site on the globe or search for it.
- Read the verdict: open (plain request), guarded (slow down), hostile (real browser), blocked (use an official API). Then read WHAT TO DO: a robots.txt ban or a login wall means use an official API, whatever the verdict color.
- Open the PLAYBOOK tab: the golden data on that kind of site, three things to build with it, and the route in. Export it as a report.
- Paste any page into X-RAY for a ranked extraction plan with proof rows and starter code.
FOR AGENTS
An MCP server and plain endpoints, free, no key.
https://rat-radar.github-a18.workers.dev/mcp
CONNECT IN ONE LINE
claude mcp add --transport http rat-radar https://rat-radar.github-a18.workers.dev/mcp
Claude, ChatGPT, Cursor, VS Code, Windsurf, OpenCode, Codex and Gemini: step by step in CONNECT YOUR AGENT at the top. No key needed.
- check_site is the site open, and how
- get_playbook what to build, with n8n, Make, Zapier and Apify recipes
- can_fetch may I fetch this URL
- xray_page how do I extract this page
- xray_diagram that page as an Excalidraw diagram
- search_sites, list_sites, summary
HOW THE DATA IS MADE
- Once a week every listed site gets one plain request to its home page, robots.txt, sitemap and llms.txt, with an honest user agent.
- A real headless browser loads the home page too. The gap between the two shows who blocks bots and who blocks only headless browsers.
- Each site gets a hostility score from 0 to 100 and a plain verdict. Protection seen by either probe counts. A robots.txt that bans every crawler, or a login wall on the home page, makes a site at least hostile.
- History is kept weekly. A change is shown only once the next weekly probe sees it again, so one bad night does not make news.
DOWNLOADS
- sites.json every site, full records
- sites.csv for spreadsheets
- summary.json headline numbers
- changes.json what changed this week
- history.json every weekly snapshot
HONEST LIMITS
- A verdict is a heuristic from one machine on one day. A site can behave differently by country, IP type or volume.
- It covers a chosen list of sites, not the internet.
- It is not legal advice. Always respect robots.txt and each site's terms, and take care with personal data.
Privacy and safety: no cookies, and what you scan is never recorded.
Rat Radar is free. Presented by RatLabs.