gautam84/appstore-intel-mcp

appstore-intel-mcp is a remote server that gives AI agents, like Claude or other assistants that use the Model Context Protocol, structured access to Google Play and the Apple App Store. Instead of an AI guessing at app information, it can call this server to search apps, pull full metadata such as titles, versions, ratings and screenshots, and fetch paginated reviews sorted by recency, rating or helpfulness. The project is built in Python with FastMCP, a framework for building Model Context Protocol servers, and it communicates over Streamable HTTP with OAuth protection. The README describes the project as early stage: version 0.1 ships three tools, search_apps, get_app_metadata and get_reviews, while a planned 0.2 release would add review analysis, competitor discovery and release tracking webhooks. Under the hood, requests flow from an MCP client through Streamable HTTP into FastMCP, which routes them through a registry to one of three tool modules for search, metadata or reviews. Those tools call separate provider modules: one uses the google-play-scraper library for Google Play, and one uses the iTunes Search API and RSS feeds for the Apple App Store. Results pass through an in-memory cache with a time limit so the same app data is not scraped again on every request. To run it locally, you clone the repository, set up a Python virtual environment, install the package, copy an example environment file, and start the server, which listens on localhost port 8000. It can then be connected to Claude Code as an MCP tool, or deployed to a hosting service called Koyeb by creating a secret key and running a deploy command, after which it can be added to Claude.ai as a custom connector using its URL and a bearer token. The roadmap lists several items not yet built: analyzing reviews for themes and sentiment with a local embedding model, comparing apps side by side, finding competitors through category and similarity matching, adding webhooks for release and rating changes, building a full OAuth 2.1 login flow for multiple users, and adding Redis so the cache can scale across more than one server instance. The project is released under the MIT license.