Summary Cupertino is a Swift-based local system for crawling, indexing, and serving Apple's developer documentation to AI agents via the Model Context Protocol (MCP). It addresses AI hallucination and enables offline development by providing an AI-ready, structured documentation archive, benefiting engineers and power users needing accurate, deterministic, and locally controlled API references.

How It Works The system crawls diverse Apple documentation sources, indexing content into a local SQLite FTS5 database employing BM25 ranking for fast, deterministic search. It then exposes this data via an MCP server, allowing AI agents like Claude to query and retrieve precise documentation. This approach prioritizes AI integration, offline access, and data ownership, built with Swift's strict concurrency and value semantics for robustness.

Quick Start & Requirements

• Primary Install: Recommended: bash <(curl -sSL https://raw.githubusercontent.com/mihaelamj/cupertino/main/install.sh) (installs binary and databases). Alternative: Homebrew (brew tap mihaelamj/tap && brew install cupertino && cupertino setup).
• Prerequisites: macOS 15+ (Sequoia). Building from source requires Swift 6.2+ and Xcode 16.0+.
• Resource Footprint: ~2-3 GB disk space for full documentation. cupertino setup takes ~30 seconds; initial full crawl takes ~12+ days.
• Links: Demo Video (YouTube), detailed command docs (docs/commands/).