Doris MCP Server is a Python-based backend service built with FastAPI that implements the Model Context Protocol (MCP). It enables clients to interact with Doris data sources through a defined set of Tools, and is designed to support NL2SQL workflows, metadata management, and analysis tasks via MCP tool calls. The server exposes two primary communication modes: a Streamable HTTP interface for high-performance web-based interactions and a Stdio interface for direct MCP client usage (e.g., Cursor). It adopts an enterprise-grade architecture centered on a Tools Manager (centralized tool registration and routing), a Resources Manager (metadata exposure), and a Prompts Manager (intelligent data analysis templates). Core database capabilities include high-performance query execution with caching, robust SQL security validation and data masking, and comprehensive metadata extraction with catalog federation across internal Doris catalogs and external sources. Enhanced by multi-worker scaling, instant security validations, hot reload of token/database configuration, and token-bound database configuration, it also ships with a web-based token management dashboard and unified configuration framework for deployment and governance in multi-tenant environments.