MCPVault
Independent directory for Model Context Protocol (MCP) servers, indexing 18,000+ servers with A-F quality grades based on weekly live verification tests.
🔗 Visit MCPVault
Description
If you're building an AI agent that needs to plug into external tools, you'll run into MCP (Model Context Protocol) servers — small connectors that let an AI model talk to a database, an API, or a piece of software. The problem is there are thousands of them, and no easy way to tell which ones are actually maintained versus abandoned. MCPVault is a directory that grades each one from A to F based on whether it actually still works.
MCPVault is an independent directory of MCP servers — the connectors that let AI models and agents interact with external tools and data sources — currently indexing over 18,000 servers. Instead of relying on static metadata, it runs a live MCP handshake verification test on each server weekly and assigns an A-F quality grade based on maintenance cadence, community adoption, and documentation quality. Server maintainers can claim and submit their listing in under two minutes, and see maintenance signals like last commit date, release cadence, and issue health. Core directory browsing, claiming, and submission are free; a paid monitoring add-on ($59/year or $7/month) adds daily handshake checks with email alerts and a public uptime status page for maintainers who want to know immediately if their server breaks.
💬 Our review
The short version: MCPVault fills a real gap for anyone working with MCP servers — a way to tell a maintained, working connector from an abandoned repo before you wire it into your agent.
Its genuine differentiator is running actual live handshake tests weekly rather than just scraping GitHub stars or last-commit dates, which is exactly the kind of signal that matters for infrastructure you're about to depend on. At this stage there's no real direct competitor doing quality-graded, verification-tested MCP discovery specifically — general options like browsing GitHub or Anthropic's own MCP docs list servers but don't grade or monitor them. The free tier covers the useful part (browsing, grades, claiming a listing) so most developers won't need to pay anything; the $59/year monitoring add-on is a reasonable price if you maintain an MCP server yourself and want to know the moment it breaks rather than finding out from a user complaint. The one thing to watch: nightly grade recalculation means a grade can lag a few hours behind a server's actual current state, and the adoption-metric methodology isn't fully documented, so treat the letter grade as a strong signal, not gospel.
💰 Pricing
📊 Global score
🤖 AI-enriched data
Annuaire, réclamation de fiche et soumission gratuits ; add-on monitoring $59/an ou $7/mois (alertes email + statut public)
Pros
Vérification live par handshake MCP réel, testée chaque semaine
Système de notation transparent, non modifiable ni monnayable
Soumission de serveur en moins de 2 minutes
Monitoring automatisé détecte les ruptures sans travail manuel
Cons
Recalcul des notes nocturne, peut retarder de quelques heures sur l'état réel
Méthodologie des métriques d'adoption pas entièrement documentée
Taux de faux positifs des tests de vérification peu documenté
Alertes quotidiennes réservées au add-on payant