simonw/browser-compat-db
Simon Willison created a GitHub repository that converts Mozilla's browser-compat-data into a ~66MB SQLite database, using an AI-generated script and GitHub Actions to build and host the database on an orphan branch with open CORS headers for easy access via Datasette Lite.
Background
- Mozilla's MDN (MDN Web Docs) is the canonical reference for web technologies. It maintains a massive dataset called **browser-compat-data** that tracks which browser versions support every CSS property, JavaScript API, HTML element, etc.
- **MCP (Model Context Protocol)** is an open standard that lets AI coding assistants (like Claude Code or Copilot) read external data sources on demand. Mozilla recently released an MCP server that lets AI tools query MDN documentation live.
- **Simon Willison** is a well-known Python developer and creator of **Datasette** (a tool for exploring SQLite databases in the browser) and **sqlite-utils** (a library for building SQLite databases).
- This project converts MDN's compatibility JSON into a single ~66MB SQLite file, then uses **GitHub Actions** to host it on a special "orphan" Git branch so the database file is served with open CORS headers — making it directly queryable from browser-based tools like **Datasette Lite**. The entire pipeline (JSON parsing → SQLite → deployment) was written by AI models (Claude Opus 4.8 and GPT-5.5).