A fully functional, self-hosted OAuth 2.0 server designed for the modern AI era and the Model Context Protocol (MCP), supporting multiple frameworks and databases, allowing users to have full control over their own data and authentication process.

The Mandoline MCP Server is an AI assistant evaluation framework that provides tools for creating custom evaluation metrics, batch scoring, and performance analysis for AI assistants such as Claude and Cursor through the Model Context Protocol, helping AI continuously improve its own performance.

Overview: MCP is a standard protocol, similar to the Type C protocol of electronic devices (which can both charge and transmit data), enabling AI models to interact with different APIs and data sources. It aims to replace fragmented Agent code integration and improve the reliability and efficiency of AI systems. By establishing a common standard, service providers can offer their own AI capabilities according to the protocol, supporting developers to build more powerful AI applications more quickly. MCP maintains the context between different applications and services, enhancing the overall ability of autonomous task execution.

MCP Marketplace is a platform that centrally displays and distributes MCP (Model Context Protocol) servers. Users can submit their own MCP servers and install them with one click through Cline. This platform aims to simplify the discovery and deployment process of MCP servers, support submission through GitHub repositories, and provide detailed installation guides.