AI agent glossary
MCP server
A program that exposes tools, data, or prompts to AI applications through the Model Context Protocol, so any compatible agent can use them.
An MCP server wraps some capability, like a database, a ticketing system, a calendar, or a set of files, and describes it in a standard way. Any MCP-compatible client can connect to it, discover what it offers, and let the model use it.
Servers can run locally on your machine, often started as a subprocess and talking over standard input and output, or remotely over HTTP, which suits shared services and hosted products. Many companies publish official MCP servers for their products, and you can write your own in a small amount of code with the official SDKs.
Treat third-party servers like any software you install. They can read data and take actions on your behalf, so use ones you trust and grant only the access they need.
A team writes a small MCP server with two tools, search_tickets and get_ticket, so every engineer's coding agent can look up bug reports without leaving the editor.
Learn more in the course
Related terms
Go from definitions to doing
Learn to use and build AI agents in 65 practical lessons.