Connect the MCP server to Claude or Cursor

In shortThe AI Rank Lab MCP server lets Claude Code, Claude Desktop, Cursor and other MCP clients call 33 AI Rank Lab tools from a conversation. Point the client at https://www.airanklab.com/api/v1/mcp and send your API key and secret as the x-api-key and x-api-secret headers. Reads of your own data are free; paid tools spend the same credits as the API, are charged up front and are refunded if they fail.

What MCP gives you

The Model Context Protocol (MCP) is an open standard that lets an AI assistant call outside tools. Once you connect AI Rank Lab, you can ask Claude or Cursor things like "Run the SEO AEO GEO check on example.com" or "Who does Google's AI Overview cite for best crm in India?" and the assistant picks the matching tool, runs it on your account and reasons over the result.

It's the same set of tools as the REST API, with the same prices and the same account. The difference is who drives: with the API you write the code, with MCP the assistant decides which calls to make. Not to be confused with the MCP command center in the dashboard, which is a Search Console chat page (lesson).

Before you start

• Copy your API key and secret from Content → Developer API → API keys (lesson).
• Use a client that can send custom headers: Claude Code, Claude Desktop and Cursor all can.
• Your MCP config stores the secret in plain text on your machine. Keep it out of version control; for a project-level Cursor config, add the file to .gitignore.

Connect your client

Claude Code

Run this in your terminal, with your own values: claude mcp add --transport http airanklab https://www.airanklab.com/api/v1/mcp --header "x-api-key: YOUR_API_KEY" --header "x-api-secret: YOUR_API_SECRET"

Claude Desktop and Cursor

Both use the same JSON block. In Claude Desktop open Settings → Developer → Edit config (the file is claude_desktop_config.json) and restart the app after saving. In Cursor, put it in ~/.cursor/mcp.json for all projects or .cursor/mcp.json for one project. Some Claude Desktop versions only load local servers from this file; if AI Rank Lab doesn't appear in the tools list after a restart, use Claude Code or Cursor instead.

{ "mcpServers": { "airanklab": { "url": "https://www.airanklab.com/api/v1/mcp", "headers": { "x-api-key": "YOUR_API_KEY", "x-api-secret": "YOUR_API_SECRET" } } } }

The full setup page with copy buttons is at /mcp, and /mcp-seo shows what agents can do with it.

Check that it works

1. Ask: "What's my AI Rank Lab credit balance?" The assistant should call the free credits tool.
2. Ask: "List my AI Rank Lab domains." That's list-domains, also free.
3. Only then try a paid tool, and say which one you want so the assistant doesn't guess.

If you get an authentication error, check that both headers are set and that you haven't rotated the key since you copied it.

The tools: free and paid

• Paid calls are charged up front and refunded automatically if the tool fails.
• `keyword-research` is metered over MCP: it reserves the maximum, charges per data lookup and returns the unused part. Over REST the same tool is a flat price. Ask for keyword-research-estimate (free) first to see the likely cost.
• `ai-citation-check` is billed per engine (ChatGPT, Claude, Gemini, Perplexity), so naming fewer engines costs less.
• Google tools take a country and device; search-locations lists the options. brand-visibility and ai-citation-check don't emulate a location.

Long-running tools and limits

Each tool call gets up to 120 seconds to finish inside the conversation. If it needs longer, the tool returns a jobId and a message instead of the result. There is no MCP tool for polling, so fetch the result over REST: GET /api/v1/jobs/<jobId> with the same two headers, or ask your assistant to do it if it can make HTTP requests. The credits were already charged and are refunded if the job fails.