The Ragie MCP Server is a Model Context Protocol (MCP) server that enables AI models to retrieve information from Ragie's knowledge base. It implements MCP and exposes a single tool, retrieve, which can be invoked by models to query Ragie's knowledge store for relevant content. The server is a Node.js/TypeScript application that requires Node.js 18 or newer and a Ragie API key (RAGIE_API_KEY) to authenticate with Ragie's API. It runs as an MCP endpoint over standard I/O, listening for protocol messages on stdin and returning results via stdout. The project uses @modelcontextprotocol/sdk for MCP integration, the ragie client to fetch knowledge items, and zod for runtime type validation. Usage involves setting RAGIE_API_KEY and running with npx @ragieai/mcp-server, with optional --partition to restrict queries and --description to override the tool description. The retrieve tool accepts query (string), topK (number, default 8), rerank (boolean, default true), and recencyBias (boolean, default false), returning an array of content chunks that match the query. The README also describes integration options for Cursor and Claude Desktop configurations.