Model Context Protocol (MCP)
Model Context Protocol (MCP) is an open standard published by Anthropic in November 2024 that defines a universal interface for connecting AI assistants to external tools, databases, and services, replacing ad-hoc per-integration code with a single protocol.
Model Context Protocol (MCP) is an open, JSON-RPC-based protocol that standardizes how large language model applications connect to external data sources and tools. Before MCP, each AI product required bespoke integration code for every external service; MCP introduces a common client-server architecture where an MCP client embedded in an AI application communicates with lightweight MCP servers that expose resources, tools, and prompt templates.
An MCP server wraps an external service—a file system, a database, a REST API, a code execution environment—and advertises its capabilities in a schema. The AI host calls the server's tool endpoints via JSON-RPC 2.0 over stdio, HTTP with Server-Sent Events, or other transports. The protocol defines three primitive types: Resources (data the model can read), Tools (functions the model can invoke), and Prompts (reusable interaction templates). Authorization, sampling, and pagination are handled at the protocol layer, not by each integration developer.
MCP matters because it lowers the cost of building and distributing AI integrations. A single MCP server for PostgreSQL works with any MCP-compatible client—Anthropic Claude, Cursor, VS Code Copilot Chat, Zed, and others—without code changes on either side. This composability mirrors the role HTTP played for web services: once a common transport existed, the ecosystem of services expanded rapidly.
By mid-2026, the MCP ecosystem includes thousands of published servers covering developer tools, CRMs, productivity suites, and data platforms. Major cloud providers including AWS and Google Cloud publish official MCP servers for their services. Competing protocols exist—notably Google's Agent2Agent (A2A) for inter-agent coordination—but MCP has become the de facto standard for tool integration in both single-agent and multi-agent pipelines.