Every MCP server, measured before you connect it.

Why we built it

Every MCP server you connect sends its tool definitions to the model on each request, can act with the access you give it, and may or may not still be maintained. Registries list servers; almost none tell you those three things. We publish them: the token cost of each server's tool list, security signals with their evidence, and a freshness rating from real repository and package activity.

Where the data comes from

• The official MCP Registry, public awesome-mcp lists on GitHub, npm and PyPI, plus submissions from maintainers.
• GitHub repository data (stars, commits, releases, license) and package registry metadata for freshness and version drift.
• Our own measurements: we connect to hosted endpoints without credentials and run local servers in an isolated sandbox. The full process is in the methodology.

Independence

MCPSkillHub is not affiliated with Anthropic, the Model Context Protocol project or the official MCP Registry. Listings are free. Scores are measurements, not endorsements, and maintainers can claim a listing to correct details or request re-analysis.

How pages are made

Every server page combines the maintainer's own published metadata (description, packages, links) with our measurements: the tool list fetched live and tokenized, security signals with evidence, and maintenance dates from GitHub and the package registries. Each page lists its sources with dates. We only ask search engines to index pages where we add something of our own: a measurement, an established project, or a maintainer-claimed listing.

Use our data

The context-tax dataset is free under CC BY 4.0. Every server has JSON at /api/v1/servers/{slug}, there is an llms.txt for AI assistants, and our own remote MCP server at https://mcpskillhub.com/mcp lets any MCP client search the registry.

Contact

Corrections, takedowns, partnerships and workshops: admin@mcpskillhub.com. See also takedown & corrections, FAQ and our LinkedIn page.