API and agent access

Endpoints

MCP server

A Model Context Protocol server over Streamable HTTP at https://airankingskool.com/mcp. It is stateless: no session id is issued and every request stands alone. GET returns 405 by design, because there is no SSE stream to attach to.

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

Tools

• search_ai_ranking_content search the guides and blog. Filter by type: topic, guide, blog, page.
• list_ai_ranking_topics the pillar topics with their URLs.
• read_ai_ranking_page any page on this domain as markdown.
• get_ai_ranking_membership what AI Ranking is, the tiers, where to join.
• check_google_rank a live Google position lookup. This is the only tool that costs money, and it is rate limited.

curl -s https://airankingskool.com/mcp \
  -H 'content-type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

A2A agent

An Agent2Agent v0.3 endpoint at https://airankingskool.com/a2a, described by /.well-known/agent-card.json. It supports blocking message/send only.

It is a retrieval agent, not a generative one. Ask it a question and it returns the published guides that answer it, with their full text. It never synthesises an answer, so it cannot invent one. Send it a URL on this domain instead of a question and it returns that page in full. Nothing is stored between requests, so tasks/get always reports the task as unknown.

Google rank check

The API behind the free Google rank checker. One request runs a live Google SERP at organic depth 100 and returns the domain's position plus the current top 10.

curl -s https://airankingskool.com/api/rank-check \
  -H 'content-type: application/json' \
  -d '{"domain":"example.com","keyword":"ai seo","location":"us","device":"desktop"}'

position is the organic rank (rank_group), the number a person means when they say "I'm number 7". It ignores People Also Ask and other SERP blocks, which is why it can read higher than tools that count every element on the page. found: false means the domain is not in the top 100 for that keyword, not that it does not rank at all.

Limits

Every uncached check is a paid upstream call, so it is capped. Five paid checks per IP per hour, then HTTP 429 with {"error":"rate_limited"}. A site-wide daily budget, then HTTP 503 with {"error":"daily_cap"}. Identical queries are cached for six hours, served free and flagged "cached": true. Both limits are temporary: back off and retry rather than rotating IPs.

Markdown for agents

Every HTML page on this site supports content negotiation. Send Accept: text/markdown and the same URL returns the article as markdown with YAML front matter, nav and footer stripped. Without that header you get ordinary HTML, so it is safe to request anywhere.

curl -s https://airankingskool.com/learn/geo/ -H 'accept: text/markdown'

Markdown responses carry x-markdown-tokens, an estimate of the body's token count (roughly four characters per token), so you can budget context before reading it.

Discovery documents

Every page also returns an RFC 8288 Link header pointing at the catalog, the OpenAPI description, this page and llms.txt, so none of the below has to be guessed at.

Content usage

robots.txt carries Content-Signal: search=yes, ai-input=yes, ai-train=no. Index this content, quote it in AI answers, cite and link back to it. Do not use it as training data for generative models.

Contact

Questions about programmatic access, or a use case that needs a higher limit: [email protected].