Model Context Protocol (MCP)
Giao thức Ngữ cảnh Mô hình (MCP)
An open standard by Anthropic that lets AI agents connect to external tools and data sources via a unified protocol.
Model Context Protocol (MCP) is an open standard introduced by Anthropic in late 2024 to standardize how AI models connect to external tools, data sources, and services. Think of it as the USB standard for AI integrations — instead of every AI app building custom connectors to every service, there's one protocol that works everywhere.
**The problem MCP solves**
Before MCP, every AI application that needed to connect to Slack, GitHub, a database, or a browser had to build a bespoke integration. This created N×M complexity — N models times M tools. MCP creates a single server/client protocol so any MCP-compatible AI can use any MCP-compatible tool.
**How it works**
MCP servers expose *resources* (readable data), *tools* (callable functions), and *prompts* (reusable templates). MCP clients (AI hosts like Claude Desktop, Claude Code, Cursor) connect to these servers and make their capabilities available to the model.
**Real-world usage**
Claude Code uses MCP extensively — it has servers for file system access, shell commands, web fetching, and more. IDE plugins like Cursor and Windsurf are adopting MCP so models can interact with the development environment without bespoke integration code.
**Ecosystem growth**
By 2025, hundreds of MCP servers had been published: GitHub, Slack, Notion, Postgres, Stripe, Puppeteer, and many more. The community-maintained list at modelcontextprotocol.io tracks them all.
**Pitfalls**
MCP servers running locally have broad system access — a malicious or compromised MCP server could read files, execute commands, or exfiltrate data. Security review of third-party MCP servers is critical before use in production.