MCP (Model Context Protocol)
Anthropic's standard for giving AI models tools and data. Like USB, but for AI connections.
MCP is a universal interface for connecting AI models to tools and data sources. Before MCP, every AI application built its own custom integration for every external system. MCP defines one standard plug: build an MCP server once and any MCP-compatible AI client can use it. A single session can connect to multiple servers simultaneously : one for your database, one for Slack, one for your file system : and the model sees all their tools as a unified toolkit. The protocol is transport-agnostic: stdio, HTTP with Server-Sent Events, or WebSockets.
How It Works
MCP servers expose tools (callable functions), resources (readable data), and prompts (templates). MCP clients connect and discover what’s available. The model calls tools by name; the client executes them and returns results into the conversation context.
Example
Quick-Fin is a 244-tool MCP server for QuickBooks Online, giving any MCP-compatible AI full access to invoices, payments, reports, and journal entries. Zero custom integration code on the client side : just connect and use. Architecture at Quick-Fin.