The Model Context Protocol (MCP) is an open standard that defines how artificial intelligence (AI) systems connect to external data sources and computational tools through a consistent client–server interface. MCP specifies that servers declare available resources and actions, while clients—such as conversational agents—can discover these capabilities and selectively use them. This design supports modular integration: one server implementation can interoperate with any MCP-compatible client, avoiding model-specific or vendor-specific plug-ins. By controlling which resources are exposed and under what conditions, MCP provides security boundaries and auditability that are essential for organizational use. In practice, the protocol enables AI systems to incorporate relevant context at runtime without uncontrolled data sharing, which improves reliability, transparency, and portability across domains.
