The Applyra MCP server: ASO data in ChatGPT, Claude or Cursor

The Applyra MCP server connects an AI assistant straight to your ASO data. Add it to ChatGPT or Claude with a URL and a sign-in, or install it in Cursor, Codex, Windsurf or VS Code with your API key, and the assistant can read the App Store and Google Play apps you track, their keywords and positions, their scores and their competitors, and track new ones. The local server is published on npm as @applyra/mcp-server, MIT licensed, with its source public and a manifest in the MCP registry.

Asking an assistant about your ASO used to mean pasting a screenshot of a table and hoping it read the numbers right. This removes that step: the assistant queries your account itself, and the same rules as the API apply to what it asks for.

Connect ChatGPT or Claude

ChatGPT and Claude connect to the hosted server, so there is nothing to install on your machine and no key to copy. Both list Applyra in their own directory. To add it by hand instead, the server URL is:

https://www.applyra.io/api/mcp

Either way the assistant then opens an Applyra page: sign in, check what it will be able to do, and allow it.

In ChatGPT

Applyra is listed in the ChatGPT plugin directory, so there is no URL to paste:

1. Open Applyra in ChatGPT, or open Plugins on chatgpt.com and search for Applyra.
2. Install the plugin.
3. On the Applyra page that opens, sign in and click Allow access.

Applyra then shows up as a plugin you can use in any conversation, to read your data and to make changes such as tracking a keyword.

You can also add it by hand: in Plugins, click Add, then Create custom MCP server. Name it Applyra, paste the URL above under Connection and keep OAuth as the authentication, then tick I understand and want to continue and click Create as a plugin.

In Claude

Applyra is listed in the Claude directory, so there is no URL to paste:

1. Open Applyra in the Claude directory, or find it in Customize, then Connectors.
2. Connect it: on the Applyra page that opens, sign in and click Allow access.

It works in Claude on the web, on desktop and on mobile, and in Claude Code. You can also add it by hand: in Customize, then Connectors, click +, then Add custom connector, and paste the URL above.

Reconnect an assistant

An assistant that still lists the Applyra tools but fails every call has lost its sign-in to Applyra. Resetting your Applyra password does this to every connected assistant, on purpose. Reconnecting takes a minute, and you keep the assistant exactly as you set it up:

• ChatGPT: open your plugin settings (Settings, then Plugins) and select Applyra. Under Connected accounts, click the ••• menu next to your account, then Reconnect. If Reconnect is not offered, choose Disconnect in the same menu, then Connect.
• Claude: open Customize, then Connectors, and select Applyra. Click Disconnect, then Connect.

Sign in to Applyra if the assistant asks you to, and its next question gets your data again.

Disconnect an assistant

The connected assistants are listed on the MCP page of the dashboard, where one click disconnects them. A local client is not on that list: it runs on your API key, so regenerating the key on the API page is what cuts it off.

Install it in a local client

Most clients use the same configuration shape. Paste this into your client's MCP config, with your own key:

{
  "mcpServers": {
    "applyra": {
      "command": "npx",
      "args": ["-y", "@applyra/mcp-server"],
      "env": { "APPLYRA_API_KEY": "YOUR_API_KEY" }
    }
  }
}

Two clients have a one line command instead:

claude mcp add applyra -e APPLYRA_API_KEY=YOUR_API_KEY -- npx -y @applyra/mcp-server
codex mcp add applyra --env APPLYRA_API_KEY=YOUR_API_KEY -- npx -y @applyra/mcp-server

The MCP page in the dashboard shows the same blocks with your key already filled in, plus the exact config file path for Claude Desktop, Cursor and VS Code.

Which clients it works with

Any client that speaks the Model Context Protocol. Today that means ChatGPT and Claude through the hosted server, and Claude Desktop, Claude Code, Cursor, Codex, Windsurf and VS Code with GitHub Copilot through the local one. The list grows on its own: nothing in the server is specific to one of them, so a new MCP client works with the URL or the configuration block above.

What the assistant can do with it

Thirty tools, covering the same ground as the API: list and add applications, list, track and untrack keywords, read rank history, inspect a keyword, read ASO Health and score history, note what you did on a day and compare before and after each event, check and simulate metadata, list, add and remove competitors, run autocomplete and niche analyses, read top charts and your account usage. They are described one by one in the MCP tools.

The prompts worth starting with are in the recipes, including the one that has an assistant draft a listing and then have Applyra score it.

What it does not do

The AI writer is deliberately not exposed. If you are already talking to a model, you do not need Applyra's model to draft your listing: you need Applyra's engine to score the draft, which simulate_metadata does. That engine is the same one behind the Metadata Simulator.

Nothing the assistant does reaches the stores either. The write tools change what your Applyra account tracks, never your published App Store or Google Play listing.

Requirements

A paid plan. ChatGPT and Claude need nothing else; a local client also needs an API key and Node 20 or later on the machine running it. The key is the one from the API. Either way the server consumes your API quota like any other client, so usage shows up in GET /api/v1/account/usage and in the API section of the dashboard. The free plan has no API access, so ChatGPT and Claude cannot be connected to a free account: see what each plan includes.