The provided text analyzes the Model Context Protocol (MCP) server registry, distinguishing between raw entries and actual distinct servers. Raw registry entries are inflated, with approximately 69% representing superseded versions of servers. After deduplication, keeping only the latest version per server name, the registry contains 31,309 distinct MCP servers. A significant portion, 22.6%, of these servers lack a source repository, making them difficult to audit. Specifically, 20.7% are remote-only servers without a repository, meaning their functionality can only be understood by connecting to them. Server hosting reveals concentration, with 17.4% of servers sharing a hostname with over 50 others, and the top three hosts serving 14.5% of all servers. The registry exhibits rapid growth, with monthly new server additions increasing dramatically over a year. This growth suggests that manual review processes will become unsustainable. The text emphasizes the importance of verifying server trustworthiness before use. It also points out that the question of which host a server is truly communicating with might be more relevant than the listed publisher. Resources are provided for checking live totals, individual server lookups, and downloading the deduplicated dataset.