Model Context Protocol (MCP)

Open standard for connecting Large Language Models (LLMs) to external data sources and executable tools, standardizing context integration.

The Model Context Protocol (MCP) is an open-source standard, introduced by Anthropic in November 2024, that solves the fragmented integration challenge for AI systems. It provides a universal interface: MCP clients (in the LLM application) communicate with MCP servers (external systems) via JSON-RPC 2.0. This protocol standardizes how LLMs access real-time data and execute functions, moving beyond static knowledge. For example, an LLM can now query a PostgreSQL database or interact with a Google Calendar through a single, secure connection. Adoption by major providers, including OpenAI and Google DeepMind, establishes MCP as the definitive bridge for transforming static LLMs into dynamic, context-aware agents.

https://modelcontextprotocol.io