Datasets:
ArXiv:
License:
Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
@@ -1,3 +1,75 @@
|
|
1 |
-
|
2 |
-
|
3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# MCPCorpus: A Large-Scale Evolvable Dataset for Model Context Protocol Ecosystem and Security Analysis
|
2 |
+
|
3 |
+
**MCPCorpus** is a comprehensive dataset for analyzing the Model Context Protocol (MCP) ecosystem, containing ~14K MCP servers and 300 MCP clients with 20+ normalized metadata attributes.
|
4 |
+
|
5 |
+
## π Dataset Overview
|
6 |
+
|
7 |
+
- **Scale**: ~14K MCP servers + 300 MCP clients
|
8 |
+
- **Attributes**: 20+ normalized metadata per artifact
|
9 |
+
- **Sources**: GitHub, community hubs, package managers
|
10 |
+
- **Applications**: Ecosystem analysis, security research, interoperability studies
|
11 |
+
|
12 |
+
## π Structure
|
13 |
+
|
14 |
+
```
|
15 |
+
MCPCorpus/
|
16 |
+
βββ Crawler/ # Data collection tools
|
17 |
+
β βββ Servers/ # Server data (mcpso_servers_cleaned.json)
|
18 |
+
β βββ Clients/ # Client data (mcpso_clients_cleaned.json)
|
19 |
+
β βββ github_info_collector.py # GitHub metadata collector
|
20 |
+
β βββ data_cleaner.py # Data normalization
|
21 |
+
βββ Website/ # Web search interface
|
22 |
+
βββ server.py # Local web server
|
23 |
+
βββ index.html # Search interface
|
24 |
+
```
|
25 |
+
|
26 |
+
## π Quick Start
|
27 |
+
|
28 |
+
### Explore Dataset
|
29 |
+
```bash
|
30 |
+
cd Website
|
31 |
+
python server.py
|
32 |
+
# Open http://localhost:8000
|
33 |
+
```
|
34 |
+
|
35 |
+
### Access Data Programmatically
|
36 |
+
```python
|
37 |
+
import json
|
38 |
+
import pandas as pd
|
39 |
+
|
40 |
+
# Load datasets
|
41 |
+
with open('Crawler/Servers/mcpso_servers_cleaned.json', 'r') as f:
|
42 |
+
servers = json.load(f)
|
43 |
+
with open('Crawler/Clients/mcpso_clients_cleaned.json', 'r') as f:
|
44 |
+
clients = json.load(f)
|
45 |
+
|
46 |
+
# Convert to DataFrame
|
47 |
+
servers_df = pd.DataFrame(servers)
|
48 |
+
clients_df = pd.DataFrame(clients)
|
49 |
+
```
|
50 |
+
|
51 |
+
### Update Dataset (Optional)
|
52 |
+
```bash
|
53 |
+
# Collect new data
|
54 |
+
cd Crawler/Servers && python Server_request.py
|
55 |
+
cd ../Clients && python Client_request.py
|
56 |
+
|
57 |
+
# Add GitHub metadata
|
58 |
+
cd .. && python github_info_collector.py --token YOUR_GITHUB_TOKEN
|
59 |
+
```
|
60 |
+
|
61 |
+
## π Citation
|
62 |
+
|
63 |
+
If you use MCPCorpus in your research, please cite it as:
|
64 |
+
```
|
65 |
+
@misc{lin2025largescaleevolvabledatasetmodel,
|
66 |
+
title={A Large-Scale Evolvable Dataset for Model Context Protocol Ecosystem and Security Analysis},
|
67 |
+
author={Zhiwei Lin and Bonan Ruan and Jiahao Liu and Weibo Zhao},
|
68 |
+
year={2025},
|
69 |
+
eprint={2506.23474},
|
70 |
+
archivePrefix={arXiv},
|
71 |
+
primaryClass={cs.CR},
|
72 |
+
url={https://arxiv.org/abs/2506.23474},
|
73 |
+
}
|
74 |
+
```
|
75 |
+
|