More SEO work now happens in an AI client: you ask Claude or ChatGPT about a competitor, an audit or a client’s site, and it calls tools to find out. An MCP server is how those tools reach the client. Keyword and backlink tools such as Ahrefs, Semrush and DataForSEO offer their own. Ours answers the questions about a site itself: what it runs on, how popular it is, whether AI search can read it, and how old it is.
The short version: connect https://checkanysite.com/mcp to your AI client, then ask about any domain in plain words. Each answer comes with its evidence: the line of robots.txt, the header, the ranked list and its date. A free key comes with monthly credits, and in the Claude app there is no key to copy.
What it answers for SEO work
| Tool | What you get |
|---|---|
get_ai_readiness |
Whether AI assistants and AI search can find, read and cite the site: a score out of 100 from its text without scripts, title and description, structured data, sitemap and llms.txt, with a fix for each gap. |
| (the same tool) | Which AI crawlers its robots.txt lets in or keeps out, such as GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot and Google-Extended. That is the site’s choice, so it’s never scored. |
get_tech_stack |
The CMS, framework, shop system, analytics and hosting the site shows, each with the line or header it was found in. |
get_popularity |
A band such as “Top 100K” and each public ranked list’s position: the Chrome UX Report and the Majestic Million. Never a visit count. |
get_domain_history |
When the domain was registered, its first TLS certificate and when the web archive first captured it. |
get_dns_records |
Where the domain points, who handles its email and what its TXT records say. |
check_website |
All of these at once, plus trust signals and what the site is. |
Things to ask
- “Which AI crawlers does nytimes.com keep out in its robots.txt? Quote the lines.”
- “How ready is my site for AI search, and what should I fix first?” (Give it your domain.)
- “Compare the tech stacks of these three competitors: which CMS and analytics does each use?”
- “Rank these five sites by popularity, and say which list each rank comes from.”
- “How old is this domain, and when did the web archive first see it?”
The agent picks the tool, and its answer cites what we found. Ask it to link our result page when a client or colleague needs to see the evidence.
Connect it
-
Claude app (web, desktop or mobile): connect CheckAnySite from the Connectors Directory and sign in. There’s no key to copy.
-
Claude Code:
claude mcp add --transport http checkanysite https://checkanysite.com/mcp \ --header "Authorization: Bearer $CHECKANYSITE_API_KEY" -
ChatGPT, Cursor, VS Code, Gemini CLI and others: the MCP docs give each client’s steps.
How to read the answers
- A stored result is the default. It answers in a moment and costs the fewest credits; its date says when we read the site. Ask the agent to scan again (
source: "scan") for a site that changed since. - Popularity is a rank, not traffic. Public lists rank sites; they don’t count visits. A band says roughly where a site stands, which is what most comparisons need. How much traffic does a website get? explains why no outside tool knows the real number.
- Blocking AI crawlers is a choice. A site that keeps GPTBot out of training may still let OAI-SearchBot in for ChatGPT’s search answers. We report each rule as written. How many websites block AI crawlers? shows what popular sites choose.
- What a site shows isn’t all it uses. A technology on the server side that leaves no trace in the page can’t be seen from outside.
What it costs
A free key comes with monthly credits, free while the API launches. Each check costs credits, fewer from a stored result than with a new scan. See every check and its price, or read Is your website ready for AI? for what the AI readiness score checks.