Model Context Protocol · Free

Use the toolkit from your AI assistant

This site is an MCP server. Connect Claude, Cursor, ChatGPT or any client that speaks the protocol, and the assistant can run the audit, validate an llms.txt, search the catalog and add a site to it — in the middle of a conversation, with no account.

Endpoint

Streamable HTTP · stateless · JSON-RPC 2.0

https://ailovemysite.com/mcp

Send POST requests with Content-Type: application/json. No session, no token: each request stands on its own. Tools that fetch a website share a budget of 30 fetches per 10 minutes per client.

Connect

Claude Code

terminal

claude mcp add --transport http ailovemysite https://ailovemysite.com/mcp

Claude Desktop, Cursor, Windsurf, VS Code

Add the server to the client's MCP configuration (a mcp.json or the app's connector settings):

mcp.json

{
  "mcpServers": {
    "ailovemysite": {
      "type": "http",
      "url": "https://ailovemysite.com/mcp"
    }
  }
}

ChatGPT and other hosted clients

Add a custom connector / remote MCP server and paste the endpoint address. Authentication: none.

Tools

audit_website read-only

Audits one page across five levels (crawlable, readable, structured, quotable, advanced) and returns the score out of 100, the level reached, every check with its status, and an ordered list of what to fix next. Results are cached for 24 hours per address.

url* · fresh ·

validate_llms_txt read-only

Fetches https://<site>/llms.txt and checks it against the llms.txt convention: Markdown, an H1 first, a blockquote summary, sections of links. Returns whether it is valid, why not, and what was found (title, summary, link and section counts, the first lines). Also reports whether llms-full.txt exists.

url* ·

search_llms_catalog read-only

Searches the catalog of websites that publish a verified llms.txt. Filter by free text, category slug and ISO country code. Returns name, site, llms.txt address, category, country and summary for each match.

query · category · country · limit · page ·

get_llms_catalog_entry read-only

Returns the catalog listing for one domain, including the beginning of its llms.txt as fetched by the server.

host* ·

list_llms_catalog_facets read-only

Lists every category and country in the llms.txt catalog with the number of listed sites in each. Use the slugs and codes with search_llms_catalog.

submit_to_llms_catalog writes

Submits a website to the catalog. The server fetches https://<site>/llms.txt itself and lists the site only if the file is a valid llms.txt. Free; no account. Category and country are detected when omitted. Re-submitting a listed site refreshes it (at most once a week).

url* · category · country · name · description ·

diagnose_page read-only

Finds what makes a page slow or messy for crawlers: heavy or unoptimised images, render-blocking scripts, redirect chains, missing compression or caching, oversized HTML, outdated tags. Returns findings by severity with a fix for each.

url* ·

find_quotable_passages read-only

Scores each passage of a page on whether it could be lifted into an AI answer as-is: self-contained, specific, under a heading. Returns the strongest and weakest passages with reasons.

url* ·

Resources: ailms://catalog/llms.txt, ailms://levels. Prompt: improve_ai_visibility.

Try it from the terminal

curl

curl -s https://ailovemysite.com/mcp \
  -H 'Content-Type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"search_llms_catalog","arguments":{"query":"shop","limit":5}}}'

For the full picture, see the MCP specification. The server speaks protocol revisions 2024-11-05 through 2025-11-25.

Why publish an MCP server?

The same reason to publish an llms.txt: assistants prefer sources they can use cleanly. A tool an assistant can call is a source it can quote with confidence, and every site added through it joins the catalog with a verified file.