Model Context Protocol (MCP)
USB for AI
8 min read
USB for AI
Before USB, every device needed its own cable and port. MCP is USB — one standard that connects AI to anything.
Before MCP, connecting an AI model to a tool required custom integration code for each combination — Claude + Notion required one integration, Claude + GitHub required another. MCP creates a universal standard: any tool that implements MCP can connect to any AI that supports MCP. One protocol, infinite connections. Cursor, Claude Desktop, and a growing ecosystem now support it.
In Plain English
MCP (Model Context Protocol) is an open standard that lets AI models connect to external tools and data sources in a standardised way. Instead of custom integrations for every tool, one protocol connects AI to databases, APIs, file systems, and applications.
The Technical Picture
MCP defines a client-server architecture where AI applications (hosts) connect to MCP servers exposing tools, resources, and prompts. Communication uses JSON-RPC 2.0 over stdio or SSE. Servers expose typed tool definitions; clients (LLMs) select and invoke tools based on context. Anthropic open-sourced the protocol in late 2024.
Real-World Examples
- Cursor using MCP to connect to your Supabase database and query it from the editor
- Claude Desktop connecting to a local filesystem MCP server to read and write files
- Hundreds of MCP servers available for Notion, GitHub, Slack, Google Drive, and more
MCP is the universal connector for AI — build one MCP server and any compatible AI model can use your tool.