An AI assistant such as Claude Code or Claude Desktop can only act on your systems through tools it has been given, and the Model Context Protocol (MCP) is the standard way to hand it those tools: a server exposes a fixed list of actions, the assistant reads the list, picks one that fits the request, calls it, and reads the structured result back. The guides in this section show how to register a server in each Claude client, where the configuration lives, and how to tell whether the connection actually works.
There are two kinds of server. A local server is a program the client starts on your machine and talks to over standard input and output; it can read files, run a database query, or call an internal API. A remote server is a URL the client talks to over HTTP; it usually authenticates you with OAuth in the browser or with a token you paste once. The commands and file paths differ between Claude Code (the terminal tool) and Claude Desktop (the chat application), and they differ again between a local and a remote server, which is why each combination has its own guide here.
HostTracker publishes a remote MCP server at mcp.host-tracker.com that exposes 65 monitoring tools over the same REST API the web app uses, so an assistant can run a live check from 300 or more locations, list what is down, pause a monitor, or post a status-page update from inside a conversation. It is a convenient first server to connect because it needs no local install and gives an immediate, visible result. See the MCP server page for the tool list and scopes.