Model Context Protocol (MCP)
MCP is an open standard by Anthropic that defines how AI models connect to external tools and data sources.
Why MCP Exists
Before MCP, every AI application built its own integration layer for tools. MCP standardizes this:
- One protocol for all tool integrations
- Tools can be shared across different AI applications
- Separation of concerns: tool providers and AI app developers work independently
MCP Architecture
| Component | Role |
|---|
| MCP Host | The AI application (e.g., Claude Desktop, your app) |
| MCP Client | Manages connection from the host to a server |
| MCP Server | Exposes tools, resources, and prompts to the model |
What MCP Servers Can Expose
- Tools — callable functions (run SQL, send email, search files)
- Resources — readable data (files, database records, API responses)
- Prompts — reusable prompt templates
Useful Phrases
- "We built an MCP server that exposes our internal knowledge base to the AI assistant."
- "MCP decouples the tool implementation from the AI application."
- "The host connects to multiple MCP servers simultaneously."