packrun

Why we included this project

Choosing a package for an AI agent to install usually ends up being a coin flip between whatever the model remembers and what looks popular at a glance. packrun replaces that guesswork with a scored comparison: it ranks npm packages across more than 50 categories by combining download trends, bundle size, maintenance activity, community signals, and TypeScript/ESM quality into a single 0-100 number. Its MCP server plugs into Cursor and Claude, giving an assistant tools to search, compare, find alternatives, and validate install compatibility before it commits to a dependency. Humans get the same data through the web interface and a public API. It's less a library and more a decision layer for package selection, useful for anyone who wants verifiable choices instead of relying on an agent's memory.