Definition
The Model Context Protocol (MCP) is an open protocol that gives AI applications a standardized interface for connecting to external data sources and tools, proposed and open-sourced by Anthropic in 2024. It is often compared to a “USB-C port” for AI applications.
How it works
MCP follows a client-server architecture: the AI application acts as the client and connects to MCP servers that expose specific capabilities. Each server wraps a file system, database or third-party service in the protocol’s common format, so an integration built once on either side can be combined with anything on the other.
Usage
Developers use MCP to let agents read local files, query enterprise databases or call SaaS services. End users typically encounter it through connectors or plugins that package ready-made MCP integrations, with no code required.
Related and confusable terms
MCP is the protocol itself. A connector is an integration that uses the protocol to reach a specific service; a plugin bundles MCP servers together with other components for distribution; and a workflow is how those tool capabilities get orchestrated into a fixed process.