Source: https://uptimeify.io/resources/platform-and-support/glossary/mcp-server

# MCP Server

> An MCP server is the far side of the [Model Context Protocol](/resources/platform-and-support/glossary/model-context-protocol): the program that offers an AI client a defined set of tools, resources and prompts.

The server holds the boundary. It declares what it offers via `tools/list`, validates every incoming argument against a JSON Schema, then runs the call against the real system behind it. The model never touches your API directly. Servers run locally over stdio or remotely over [Streamable HTTP](/resources/platform-and-support/glossary/streamable-http), where OAuth handles authorization and the [API token](/resources/platform-and-support/glossary/api-token) scope decides what a client is allowed to see. The [Uptimeify MCP server](/mcp-server) is remote: monitors, incidents and uptime summaries plus diagnostics for DNS, TLS, WHOIS and ports, always within the scope of the API token you connect with.