Model Context Protocol (MCP)
Learn about the Model Context Protocol for connecting AI models to external tools and data sources securely. What is MCP? The Model Context Protocol (MCP) is an open standard for connecting AI models with external tools, data sources, and services. It defines how models discover available tools, make requests, and handle responses securely. MCP provides a standardized interface similar to how USB standardizes device connections. MCP Architecture MCP has three component types: Servers (provide tools and resources), Clients (connect to servers and request actions), and Hosts (applications that run models). A model can access multiple MCP servers, each providing different capabilities: database access, file system, web search, API integration, code execution. Security Model MCP uses OAuth 2.0 for authentication. Tools define their own access scopes. Users approve tool access in a UI layer before the model can use them. All tool calls are logged. Sensitive operations require explicit user approval. This prevents the model from accessing capabilities the user hasn't authorized.