It remembers, because it tracks
Most SEO MCP servers query data live, one question at a time. LemRank checks your keywords every day, so the assistant can answer what changed since last week rather than only what is true right now.
Model Context Protocol
Give Claude, ChatGPT or Cursor direct access to your Google positions. Ask what moved, what is cannibalising what, and what to track next — answered from your own daily rank data, not from a live guess.
Install
/api-docs) and use the API Keys tab. The key is shown once and stored only as a hash.YOUR_LEMRANK_API_KEY.https://lemrank.com/api/public/mcp
{
"mcpServers": {
"lemrank": {
"url": "https://lemrank.com/api/public/mcp",
"headers": {
"Authorization": "Bearer YOUR_LEMRANK_API_KEY"
}
}
}
}claude mcp add --transport http lemrank https://lemrank.com/api/public/mcp \ --header "Authorization: Bearer YOUR_LEMRANK_API_KEY"
In ChatGPT, add it as a remote connector using the same endpoint and put the key in an Authorization: Bearer header. Any client that speaks Streamable HTTP MCP will work — the server holds no session state. The same data is available over REST.
Tools
| Tool | What it returns |
|---|---|
| list_projects | Every site the API key tracks, with its domain, tracking frequency, default language and default location. |
| get_project | One tracked site, checked against the key's owner. |
| list_keywords | Current Google positions, previous and best position, ranking URL, volume, CPC and difficulty. |
| ranking_history | Up to 90 days of daily positions, so the assistant can describe movement rather than guess. |
| serp_snapshots | Whether an AI Overview, featured snippet or local pack was present when the result page was captured. |
| track_keyword | Add a keyword to tracking. Returns the existing one instead of creating duplicates. |
| url_intel | For one URL you own: ranking keywords, 30-day sparklines, cannibalisation, toxic backlink count and AI citations. |
| whoami | Confirm which account the configured key belongs to. |
What it does not do: it has no site crawler and no third-party backlink index of its own. For a technical crawl, pair it with a crawler MCP — we say which ones in the roundup.
Why this one
Most SEO MCP servers query data live, one question at a time. LemRank checks your keywords every day, so the assistant can answer what changed since last week rather than only what is true right now.
Daily positions come back as a series, which is what an assistant needs to spot a slide before it becomes a crisis.
Every call carries your own API key and is re-checked against the account that owns each project. A key can never read another workspace.
It is a remote server over Streamable HTTP. Nothing to run on your machine, nothing to keep awake for a scheduled task.
In practice
Check it works
curl -X POST https://lemrank.com/api/public/mcp \
-H "Content-Type: application/json" \
-H "Accept: application/json, text/event-stream" \
-H "Authorization: Bearer YOUR_LEMRANK_API_KEY" \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'Rate limits and error codes match the REST API: 60 requests a minute per key, and a clear message when a key is missing, expired or out of quota.
The MCP server reads whatever LemRank is tracking for you, so the first step is adding your keywords. The trial is free and needs no card.