MCP server
A program that exposes tools, resources, or prompts to an AI application over the Model Context Protocol.
An MCP server is the component that actually does things: reads a database, calls an API, runs a shell command, searches a filesystem. It advertises its capabilities to the client during a handshake, then responds to invocation requests. Servers can run locally as a subprocess (communicating over stdio) or remotely as a networked service (communicating over HTTP).
Why it matters
A server is a trust boundary. The text it returns — tool descriptions, tool output, resource contents — is read by a language model that may act on it. A server can be malicious by design, compromised after the fact, or simply careless with the third-party data it relays. Remotely reachable servers are also exposed infrastructure: if one is deployed without authentication in front of it, anyone who finds it can use whatever the server can do.
Exposure map
Live counts of instances showing this pattern will appear here once the exposure map is collecting data.