RepoAI is a directory for MCP (Model Context Protocol) servers — the plugins connecting Claude, Cursor, VS Code, and other AI tools to your files, databases, and APIs. Every one of 213 listed servers gets a 0-100 security score from 15+ signals, a ready-to-paste config for 5 AI clients, and a verdict written by a human editor — never scraped, never AI marketing copy. Browse 36 categories or jump straight to Best MCP servers for your client. No paid placements, ever.

Hey Product Hunt 👋 I built RepoAI because every time I wanted to add an MCP server to Claude Desktop or Cursor, I'd end up with 15 browser tabs open trying to figure out: is this repo actually maintained? Does it need my API keys? Is anyone still watching this thing for security issues? MCP directories exist, but most just mirror GitHub — stars, a README, done. None of them tell you the one thing that actually matters before you install something that gets tool-call access to your machine: how risky is it? So RepoAI scores every server 0-100 across 15+ signals — auth method, read-only mode, maintenance activity, license, and more — and pairs that with a human-written verdict (never auto-published AI copy) and a config you can paste straight into your client of choice. Still early — 213 servers across 36 categories, growing every week. Next up: extending the same trust model beyond MCP to Claude Skills and other AI-ecosystem tools. Would genuinely love feedback, especially from anyone who's been burned by installing a sketchy MCP server. What would make you trust a listing more?

Love the security scoring approach, that's been a real concern with MCP servers and something most directories ignore. One thing that would help me out: a "tested on date" timestamp next to each server entry, since these things break fast when the underlying APIs change and I want to know if the config snippet I'm copying will actually still work.

RepoAI.io

The security-scored directory for MCP servers

The security-scored directory for MCP servers