What is the Model Context Protocol (MCP)?

The Model Context Protocol (MCP) is an open standard, introduced by Anthropic in late 2024, that lets AI agents discover and call external tools through a JSON-RPC interface. It's the USB-C of agentic AI — instead of writing custom integrations for every model and tool combination, you write one MCP server and every MCP-compatible agent (Claude Desktop, Cursor, Cline, Zed, Continue.dev, plus LangChain and mcp-agent runtime wrappers) can use it. As of mid-2026 the Anthropic MCP Registry lists over 1,200 servers spanning databases (Postgres, SQLite, MongoDB), filesystems, GitHub, Slack, web scraping (AgentFetch, mcp-firecrawl), search, and bespoke SaaS APIs. The protocol defines three primitives — Tools (callable functions), Resources (readable data), and Prompts (parameterized templates) — and transports over stdio, SSE, or streamable-HTTP. For developers building agents, MCP collapses what used to be N×M integration work (every agent × every tool) into N+M (each side speaks one protocol). For end users, it means installing capabilities into Claude Desktop or Cursor is as simple as pasting a JSON config snippet.