Add the Storelift MCP server to Cursor

How this guide was checked

Cursor's documentation: cursor.com.

Version

This guide describes the server that is live: storelift-mcp 1.1.1 on npm (2026-09-14) and the hosted endpoint https://storelift.net/mcp, both with the same 8 tools: get_ai_visibility, get_charts, get_history, get_keywords, get_reviews, get_rivals, get_store_page, list_apps.

Every tool in this release reads; none of them changes anything in your account. The local server answers over stdio with protocol version 2024-11-05, the hosted endpoint over Streamable HTTP with 2025-06-18; the endpoint is stateless, with no session id and no server-sent stream.

Requirements

• Plan. MCP access needs the Pro, Studio or Enterprise plan, both with an API key and with sign-in (OAuth).
• Key. Storelift → Settings → API keys. The key starts with sl_live_. Keys in this release carry no scopes: a key reads every app in its workspace. Keep it only in the client's configuration.
• Runtime. Node.js 18 or later for the local server (npx fetches the package on first run). The hosted endpoint needs nothing installed.
• Cursor. The editor and the Cursor CLI read the same mcp.json: .cursor/mcp.json in a project, or ~/.cursor/mcp.json for every project.

Setup

Local server with an API key — add this to .cursor/mcp.json:

{
  "mcpServers": {
    "storelift": {
      "command": "npx",
      "args": [
        "-y",
        "storelift-mcp"
      ],
      "env": {
        "STORELIFT_API_KEY": "sl_live_..."
      }
    }
  }
}

Or the hosted endpoint with sign-in (OAuth): Cursor opens the sign-in in your browser for servers that ask for it.

{
  "mcpServers": {
    "storelift": {
      "url": "https://storelift.net/mcp"
    }
  }
}

Cursor asks for approval before using MCP tools by default. From the command line, cursor-agent mcp list-tools storelift checks the connection and lists the tools with their arguments.

First read

Ask: “List my Storelift apps, then show the keyword ranks for the first one in the US.” The assistant calls list_apps first, then get_keywords for the app you pick. The list_apps reply looks like this (Storelift sample data, 2026-09-28; your account returns your own apps):

{
  "apps": [
    {
      "id": "sbx-lumen",
      "name": "Lumen — Sleep Sounds",
      "countries": [
        "us",
        "gb",
        "de"
      ],
      "keywords": [
        "sleep sounds",
        "white noise",
        "rain sounds"
      ]
    }
  ]
}

get_keywords returns one row per tracked keyword:

{
  "country": "us",
  "platform": "ios",
  "keywords": [
    {
      "term": "sleep sounds",
      "measured": true,
      "rank": null,
      "demand": 73,
      "searchDepth": 200
    },
    {
      "term": "white noise",
      "measured": true,
      "rank": 57,
      "demand": 37,
      "searchDepth": 200
    }
  ]
}

A number in rank is the rank; null means the keyword was measured and the app is not in the results; measured: false means it could not be measured. The three stay apart: counting an unmeasured day as zero gives a wrong answer.

The connection works. Next, three first tasks: which keywords fell, who is ahead on them, and what to say when a keyword has no measurement today, each with the answer to expect.

Removing access

1. In Cursor, turn the server off or remove it under Customize, or delete the storelift entry from mcp.json. From the CLI, cursor-agent mcp disable storelift stops it from loading.
2. Delete the key under Storelift → Settings → API keys. The next request with that key is refused (a tool error from the local server, HTTP 401 from the hosted endpoint); there is no open session to wait out.
3. If you connected with sign-in (OAuth), remove the connection under Settings → API → Connected apps; that connection's token gets a 401 on its next request.

Troubleshooting

Read next