Skip to content

MCP (Model Context Protocol)

← All terms · AI agents and harnesses

Also called MCP Servers

An open standard protocol that enables AI models to securely connect to and interact with external custom tools, local filesystems, and enterprise data sources.

What it is

Created by Anthropic, MCP acts like a "USB-C port" for AI agents. By defining a universal interface, developers can write an "MCP Server" once and allow any compatible agent (like Cline, Windsurf, or Claude Code) to instantly discover and use those specific tools. This replaces fragmented, one-off API integrations.

When you would use it

You implement an MCP server when you need to extend an AI agent's capabilities beyond its default toolset, safely hooking it into your organization's unique backend systems or data silos.

Common operations

  • Connecting an AI agent to a local database to execute queries securely.
  • Allowing an IDE agent to read live corporate documentation directly from internal wikis.
  • Granting an agent secure, sandboxed access to cloud infrastructure commands.

Related terms

Where this is taught

Going deeper