How to Connect CiteTrack AI to Claude, ChatGPT, and n8n

CiteTrack tracks your AI visibility inside WordPress. But sometimes you want that data somewhere else. In a Claude conversation. In a ChatGPT workflow. In an n8n flow that posts a weekly digest to Slack. Or in your own script.

That's what the developer API and MCP are for. Let's wire it up.

What you can connect to

Every CiteTrack site exposes two things, and both are token-authed so nothing is public:

• An MCP server at /wp-json/citetrack/v1/mcp. This is what Claude, ChatGPT, n8n, Make, and code connect to. It speaks the Model Context Protocol, so an AI assistant can call your data as tools.
• A plain REST API at /wp-json/citetrack/v1/…, with read endpoints like /snapshot, /citations/share-of-voice, /health, and /crawl-log. Handy if you'd rather just curl it from code.

The data never leaves your server except to answer your own request. The token acts on behalf of your admin account, so treat it like a password.

Step 1 — Generate your access token

In wp-admin, go to CiteTrack AI → API & MCP and click Generate access token. Copy it right away, because it's shown only once (it starts with ctmcp_). If you lose it, just rotate it and update wherever you pasted it.

That screen also shows your exact MCP URL and REST base, so keep it open for the next step.

Step 2 — Connect your tool

Here are the three most common setups.

Claude Desktop

Claude Desktop reads a config file, and that's the reliable way in. Go to Settings → Developer → Edit Config, and add an mcpServers block inside your existing config object:

{
  "mcpServers": {
    "citetrack": {
      "type": "http",
      "url": "https://yoursite.com/wp-json/citetrack/v1/mcp",
      "headers": { "Authorization": "Bearer YOUR_TOKEN" }
    }
  }
}

Replace the URL and YOUR_TOKEN, then fully quit and reopen Claude Desktop. If your Claude build doesn't support remote http servers, use the bridge instead (it needs Node.js):

npx -y mcp-remote https://yoursite.com/wp-json/citetrack/v1/mcp --header "Authorization: Bearer YOUR_TOKEN"

ChatGPT, n8n, and Make

Same MCP URL, same Bearer header: Authorization: Bearer YOURTOKEN. Where a tool only accepts a URL and no headers, append ?token=YOURTOKEN to the URL instead (note the ?). That's it.

Plain REST from code

Prefer to skip MCP entirely? Every read endpoint also works with a WordPress Application Password (Users → Profile → Application Passwords) over HTTPS Basic auth:

curl https://yoursite.com/wp-json/citetrack/v1/snapshot 
  --user "USERNAME:APPLICATION_PASSWORD"

That returns your crawl health, share of voice, and recent reputation changes as JSON, ready to drop into a dashboard or a cron job.

Step 3 — Ask it something

Once it's connected, you don't memorize endpoints. You just ask. The MCP exposes a clean set of tools, and the assistant picks the right ones:

• get_snapshot for the overall picture in one call.
• list_citations for cited, mentioned, or absent on each prompt and engine.
• shareofvoice, you versus your named competitors.
• indexednotcited, the pages AI crawls a lot but never quotes, which is your best fix list.
• reputation_changes for dropped citations, sentiment shifts, and rewritten answers.
• crawlhealth and recentcrawls for your AI crawler activity.
• list_prompts, everything you're tracking.

So you can ask Claude, "Using CiteTrack, what's my share of voice this week and which pages are crawled but never cited?" and get a real answer built from your own data. Or have n8n call get_snapshot every Monday and post the summary to Slack.

What it won't do (yet)

Two honest limits, so you don't go hunting:

• No API cost data in the MCP. The tools cover visibility, citations, crawl, and reputation. Your API spend lives on the Insights → Runs & cost screen in wp-admin, not in the API. Track cost there and with budget caps at each provider.
• run_citations spends real credit. There's a tool to trigger a citation run on demand, and it uses your own API key just like a scheduled run. Handy, but it's not free, so use it deliberately.

Where this gets powerful

The single-site setup is useful on its own. But the real payoff is across many sites. If you run an agency, you can connect every client's MCP and pull one combined report from a single Claude conversation or script. We walk through exactly that, including a copy-paste portfolio script, in how to manage multiple clients with CiteTrack AI.

Frequently asked questions

Related reading

• How to Manage Multiple Clients with CiteTrack AI
• How to Track Whether AI Is Citing Your Site (AEO Guide)
• Best AI Visibility Tools for WordPress
• 9 Best Tools to Track AI Citations

Guides · API, Claude, How-To, MCP