Ask your AI app whether AI assistants recommend your business — and who they name instead. This MCP server brings SeekinWeb into any MCP client, without leaving the tool.
Does an AI assistant with web search recommend the business? Who is named instead, which sites it relies on, and what it says about it.
AI Visibility Score 0-100 across 8 weighted signals, with a concrete fix per failing check.
A ready-to-publish llms.txt (llmstxt.org) derived from the real page content.
Open-source and dependency-free. Clone the repo and point your MCP client at mcp.js:
git clone https://github.com/draxork/seekinweb.com
Then add to your MCP client config (e.g. Claude Desktop’s claude_desktop_config.json):
{
"mcpServers": {
"seekinweb": {
"command": "node",
"args": ["/ruta/a/seekinweb.com/mcp.js"]
}
}
}
Node 18+. Prefer the website? Scan any domain free at seekinweb.com.
The easiest way: add it as a custom connector in your AI app (remote MCP, no sign-in) with this URL. The same tools are exposed over Streamable HTTP as JSON-RPC 2.0 (initialize, tools/list, tools/call):
POST https://seekinweb.com/api/mcp
Try it from a terminal:
curl -s https://seekinweb.com/api/mcp \
-H 'Content-Type: application/json' \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/call",
"params":{"name":"scan_website","arguments":{"url":"example.com"}}}'
Read-only and rate-limited. tools/call runs a real scan, same scope as the free web scanner.
An MCP (Model Context Protocol) server that exposes SeekinWeb's AI-visibility scanner as tools an AI agent can call directly: scan_website and generate_llms_txt. Any MCP-compatible client (Claude Desktop, Cursor, Cline) can use it.
check_ai_recommendations(url) shows whether an AI assistant with web search recommends the business, who it names instead, which sites it relies on and what it says about it. scan_website(url) returns a 0-100 AI Visibility Score across 8 weighted signals with a concrete fix for each failure. generate_llms_txt(url) returns a ready-to-publish llms.txt derived from the page.
No. The scan is free and needs no signup, whether you use the website or the MCP server.
Clone the open-source repo, then point your MCP client config at mcp.js. No npm install, no dependencies — it runs on Node 18+ with the standard library only.
Yes. A remote HTTP endpoint at /api/mcp speaks the same JSON-RPC 2.0 (initialize, tools/list, tools/call). Point an HTTP-capable MCP client at it, or call it with curl — no clone, no install. It is read-only and rate-limited.