MCPCorpus 是一个全面的数据集,用于分析模型上下文协议(MCP)生态系统,包含约14K个MCP服务器和300个MCP客户端,每个实体具有20多个标准化元数据属性。
MCPCorpus/
├── Crawler/ # 数据收集工具
│ ├── Servers/ # 服务器数据
│ ├── Clients/ # 客户端数据
│ ├── github_info_collector.py # GitHub元数据收集器
| ├── data_cleaner.py # 数据规范化
│ └── tool_extractor.py # MCP工具提取
└── Website/ # 网页搜索界面
├── server.py # 本地Web服务器
└── index.html # 搜索界面
└── mcpso_servers_cleaned.json
└── mcpso_clients_cleaned.json
cd Website
python server.py
# 打开 http://localhost:8000
import json
import pandas as pd
# 加载数据集
with open('Crawler/Servers/mcpso_servers_cleaned.json', 'r') as f:
servers = json.load(f)
with open('Crawler/Clients/mcpso_clients_cleaned.json', 'r') as f:
clients = json.load(f)
# 转换为DataFrame
servers_df = pd.DataFrame(servers)
clients_df = pd.DataFrame(clients)
# 收集新数据
cd Crawler/Servers && python Server_request.py
cd ../Clients && python Client_request.py
# 添加GitHub元数据
cd .. && python github_info_collector.py --token YOUR_GITHUB_TOKEN
如果您在研究中使用了MCPCorpus,请引用如下:
@misc{lin2025largescaleevolvabledatasetmodel,
title={A Large-Scale Evolvable Dataset for Model Context Protocol Ecosystem and Security Analysis},
author={Zhiwei Lin and Bonan Ruan and Jiahao Liu and Weibo Zhao},
year={225},
eprint={2506.23474},
archivePrefix={arXiv},
primaryClass={cs.CR},
url={https://arxiv.org/abs/2506.23474},
}