How the catalog grew to 80,000+ servers: nine sources, one index
Four new ingest sources, live ecosystem stats at /stats, a fresh-servers feed at /new with RSS, and hourly health checks for remote MCPs.
The MCP ecosystem does not live in one registry, so neither do we. The catalog now ingests from nine sources: the official MCP Registry, npm, PyPI, crates.io, Docker Hub, GitHub search, Glama, PulseMCP and Smithery. After deduplication that is 80,000+ servers in one searchable index.
Watch it live
- /stats shows the ecosystem in real time: catalog size, source and category breakdowns.
- /new lists servers as they appear, with an RSS feed if you want new MCPs in your reader.
- /browse got sorting: by installs, newest, rating or name.
Quality over quantity
A big index is only useful if it is honest. Remote servers get an hourly health check, and the status is visible on every server page. Server descriptions that arrive in Chinese, Korean or Japanese are translated to English so search works across the whole index. Each server also carries a security snapshot: whether the code is public, what credentials it asks for.
Nine sources disagree about names, versions and ownership constantly. Resolving that mess into one clean page per server is most of the work, and it is exactly the part you never have to think about.