NOTE This article explains the concept and background of MCP. For the configuration reference, see MCP server.
What is MCP?
MCP (Model Context Protocol) is an open protocol. It allows AI systems to communicate with external tools, apps, and data sources. The communication is standardized and secure.
MCP was released by Anthropic in late 2024 as an open standard.
MCP works like a translator or bridge. An AI model can use it to retrieve and process information from the outside world. Examples include reading a calendar, querying a database, or fetching live data from an API.
Without MCP, each AI model would need its own custom integration. That quickly leads to inconsistencies and increased maintenance effort. MCP instead defines a common language and structure for these connections. Any compliant tool can connect to any compliant AI system.
MCP provides mechanisms for privacy and access control. You decide which tools AI systems can access and which data they can see.
In short, MCP enables more capable and flexible connections with AI systems. At the same time, security, consistency, and maintenance remain easy to manage.
Why MCP server instead of REST API?
The 'API' (RestApiDefinition) entity type supports two different classes:
MCP server (
McpRestApi): provides tools for an AI system. Based on the MCP protocol.REST API (
RestApi): defines REST API endpoints for other integration purposes.
Both classes share the same input form. Two tabs are class-specific:
'API' tab: available for both classes, with class-specific content in each.
'Meta information' tab: available exclusively for the 'REST API' class.
The starting point for managing both classes is the menu item API Manager.
The 'MCP server' class is optimized for connecting AI systems via the MCP protocol. The 'REST API' class is used for classic REST interface definitions.
See also: Configuration reference for the MCP server: MCP server.