HumanMCP：用于评估MCP工具检索性能的人类查询数据集

Model Context Protocol (MCP) servers host thousands of open-source standardized tools that connect large language models (LLMs) to external systems. However, existing datasets and benchmarks lack realistic, human-like user queries, creating a significant gap in evaluating the tool usage and ecosystems of MCP servers. While current datasets include tool descriptions, they fail to accurately represent the diverse ways different users articulate their requests, resulting in poor generalization and inflated reliability of certain benchmarks. This research introduces the first large-scale MCP dataset featuring a variety of high-quality user queries specifically generated to correspond with 2800 tools across 308 MCP servers, building upon the MCP Zero dataset. Each tool is associated with multiple user queries to enhance the dataset's diversity and applicability.