Content
# chinadrugtrials-mcp
MCP Server for China Drug Clinical Trial Registration and Information Disclosure Platform (药物临床试验登记与信息公示平台, chinadrugtrials.org.cn).
## Overview
Provides tools to search and retrieve clinical trial data registered with China's NMPA/CDE.
Inspired by [clinicaltrialsgov-mcp-server](https://github.com/cyanheads/clinicaltrialsgov-mcp-server) architecture.
## Capabilities
### Tools (3)
| Tool | Description |
|------|-------------|
| `chinadrugtrials_search_trials` | Search clinical trials by drug name, indication, registration number, sponsor, PI, trial site, status, drug type |
| `chinadrugtrials_get_trial_detail` | Get detailed information for a specific trial by its registration number (e.g., CTR20261395) |
| `chinadrugtrials_get_statistics` | View clinical trial registration statistics from the platform |
### Resources (1)
| Resource | URI Pattern | Description |
|----------|-------------|-------------|
| Trial Detail | `chinadrugtrials://{reg_no}` | Direct access to a single trial by CTR registration number |
### Prompts (1)
| Prompt | Description |
|--------|-------------|
| `analyze_china_trial_landscape` | Guided analysis of the China clinical trial landscape for a given therapeutic area or drug class |
## Architecture
```
chinadrugtrials-mcp/
├── server.py # Entry point (FastMCP initialization)
├── chinadrugtrials_mcp/
│ ├── __init__.py
│ ├── constants.py # URLs, recovery hints
│ ├── schemas.py # Pydantic input/output models
│ ├── browser_manager.py # Playwright browser lifecycle
│ ├── parsers.py # HTML parsing + formatting utilities
│ ├── services.py # Core service layer (HTTP/browser ops)
│ ├── tools.py # Tool handler implementations
│ ├── resources.py # Resource handler implementations
│ └── prompts.py # Prompt template implementations
├── pyproject.toml
├── requirements.txt
└── README.md
```
## Technical Notes
- Uses Playwright for browser automation due to JavaScript anti-bot protection on the target site.
- The site renders search results as full HTML pages (server-side rendering after JS anti-bot challenge).
- Trial detail pages use session-based POST requests, so the full browser context is needed.
- Service layer provides retry-friendly error handling with recovery hints.
## Installation
```bash
# Install package (editable)
pip install -e .
# Or install dependencies manually
pip install mcp httpx beautifulsoup4 lxml playwright pydantic
# Install Playwright browser
playwright install chromium
```
## Configuration (WorkBuddy MCP)
Add to `~/.workbuddy/mcp.json`:
```json
{
"mcpServers": {
"chinadrugtrials": {
"type": "stdio",
"command": "/path/to/venv/bin/python",
"args": ["/path/to/chinadrugtrials-mcp/server.py"]
}
}
}
```
## Registration Number Format
China drug trial registration numbers follow the pattern: `CTR{YYYY}{NNNN}`
Examples: `CTR20261395`, `CTR20240001`
