| name | mcp-server-12306 |
| description | MCP server for querying China Railway 12306 train tickets, stations, transfers, and schedules in real-time |
| triggers | ["how do I query train tickets in China","set up 12306 MCP server for train information","search for railway stations by name or pinyin","find train transfer connections between cities","check train ticket prices and availability","get train route and schedule information","query China railway data through MCP","integrate 12306 train queries with AI assistant"] |
MCP Server 12306
Skill by ara.so — MCP Skills collection.
MCP Server 12306 is a Model Context Protocol server that provides real-time access to China Railway 12306 data including train tickets, station information, transfers, prices, and schedules. Built with FastAPI for high-performance async operations.
What It Does
This MCP server exposes tools for:
- Real-time ticket queries: Search available tickets with seat types, departure/arrival times
- Station search: Fuzzy search for stations by Chinese name, pinyin, or abbreviation
- Transfer planning: Find one-transfer routes between cities
- Price queries: Get real-time ticket prices for trains
- Route information: Query train stops and schedules
- Time utilities: Get current time and relative dates for booking
Installation
For Claude Desktop (Stdio Mode - Recommended)
Add to your Claude Desktop configuration file:
macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json
Using uvx (fastest):
{
"mcpServers": {
"12306": {
"command": "uvx",
"args": ["mcp-server-12306"]
}
}
}
Using pipx:
{
"mcpServers": {
"12306": {
"command": "pipx",
"args": ["run", "--no-cache", "mcp-server-12306"]
}
}
}
From source (for development):
{
"mcpServers": {
"12306": {
"command": "uv",
"args": ["run", "python", "-m", "mcp_12306.cli"],
"cwd": "/path/to/mcp-server-12306"
}
}
}
For Remote/HTTP Mode (Streamable HTTP)
Local development:
git clone https://github.com/drfccv/mcp-server-12306.git
cd mcp-server-12306
uv sync
uv run python scripts/start_server.py
Then configure your MCP client:
{
"mcpServers": {
"12306": {
"url": "http://localhost:8000/mcp"
}
}
}
Docker deployment:
docker run -d -p 8000:8000 --name mcp-server-12306 drfccv/mcp-server-12306:latest
Available Tools
1. query_tickets
Search for available train tickets between stations.
Parameters:
from_station (string, required): Departure station name (Chinese or pinyin)
to_station (string, required): Arrival station name (Chinese or pinyin)
train_date (string, required): Travel date in YYYY-MM-DD format
purpose_codes (string, optional): "ADULT" (default) or "0X00" for student tickets
Example usage:
result = await mcp_client.call_tool("query_tickets", {
"from_station": "北京",
"to_station": "上海",
"train_date": "2025-06-15",
"purpose_codes": "ADULT"
})
Response includes:
- Train number, type (G/D/K/T/Z)
- Departure/arrival times and duration
- Seat availability and types
- Station codes
2. query_ticket_price
Get real-time ticket prices for a specific train.
Parameters:
train_no (string, required): Train number (e.g., "G101")
from_station_no (string, required): Departure station code
to_station_no (string, required): Arrival station code
seat_types (string, required): Comma-separated seat type codes
train_date (string, required): Travel date in YYYY-MM-DD format
Example usage:
result = await mcp_client.call_tool("query_ticket_price", {
"train_no": "G101",
"from_station_no": "BJP",
"to_station_no": "SHH",
"seat_types": "9,M,O",
"train_date": "2025-06-15"
})
Response includes:
- Price for each seat type
- Seat type codes and names
- Currency (CNY)
3. search_stations
Search for railway stations by name, pinyin, or abbreviation.
Parameters:
keyword (string, required): Search term (Chinese, pinyin, or abbreviation)
Example usage:
result = await mcp_client.call_tool("search_stations", {
"keyword": "beijing"
})
result = await mcp_client.call_tool("search_stations", {
"keyword": "北京"
})
Response includes:
- Station name (Chinese)
- Station code
- Pinyin and abbreviation
- Match score
4. get_station_info
Get detailed information about a specific station.
Parameters:
station_name (string, required): Exact station name or code
Example usage:
result = await mcp_client.call_tool("get_station_info", {
"station_name": "北京南"
})
5. query_transfer
Find one-transfer routes between cities (when no direct train available).
Parameters:
from_station (string, required): Departure station
to_station (string, required): Arrival station
train_date (string, required): Travel date in YYYY-MM-DD format
purpose_codes (string, optional): Ticket type
Example usage:
result = await mcp_client.call_tool("query_transfer", {
"from_station": "北京",
"to_station": "桂林",
"train_date": "2025-06-20",
"purpose_codes": "ADULT"
})
Response includes:
- First leg train details
- Transfer station
- Second leg train details
- Total duration
- Layover time at transfer station
6. get_train_route_stations
Query all stops and schedule for a specific train.
Parameters:
train_no (string, required): Train number
from_station_telecode (string, required): Origin station code
to_station_telecode (string, required): Destination station code
depart_date (string, required): Departure date in YYYY-MM-DD format
Example usage:
result = await mcp_client.call_tool("get_train_route_stations", {
"train_no": "G101",
"from_station_telecode": "BJP",
"to_station_telecode": "SHH",
"depart_date": "2025-06-15"
})
Response includes:
- Station sequence
- Arrival/departure times
- Stop duration
- Running distance
7. get_current_time
Get current time and calculate relative dates for booking.
Parameters:
timezone (string, optional): Timezone (default: "Asia/Shanghai")
Example usage:
result = await mcp_client.call_tool("get_current_time", {
"timezone": "Asia/Shanghai"
})
Response includes:
- Current time
- ISO format timestamp
- Unix timestamp
- Formatted date strings
- Relative dates (today, tomorrow, next week)
Common Patterns
Complete ticket search workflow:
time_info = await client.call_tool("get_current_time", {})
travel_date = time_info["dates"]["tomorrow"]
stations = await client.call_tool("search_stations", {
"keyword": "beijing"
})
from_code = stations[0]["code"]
stations = await client.call_tool("search_stations", {
"keyword": "shanghai"
})
to_code = stations[0]["code"]
tickets = await client.call_tool("query_tickets", {
"from_station": "北京",
"to_station": "上海",
"train_date": travel_date,
"purpose_codes": "ADULT"
})
if tickets["trains"]:
train = tickets["trains"][0]
prices = await client.call_tool("query_ticket_price", {
"train_no": train["station_train_code"],
"from_station_no": train["from_station_telecode"],
"to_station_no": train["to_station_telecode"],
"seat_types": train["seat_types"],
"train_date": travel_date
})
Finding transfer routes:
transfers = await client.call_tool("query_transfer", {
"from_station": "拉萨",
"to_station": "厦门",
"train_date": "2025-07-01",
"purpose_codes": "ADULT"
})
for route in transfers["transfer_routes"]:
print(f"Transfer at: {route['transfer_station']}")
print(f"Layover: {route['layover_minutes']} minutes")
print(f"Total duration: {route['total_duration']}")
Checking train route details:
route = await client.call_tool("get_train_route_stations", {
"train_no": "G7",
"from_station_telecode": "BJP",
"to_station_telecode": "SHH",
"depart_date": "2025-06-15"
})
for stop in route["stations"]:
print(f"{stop['station_name']}: {stop['arrive_time']} - {stop['start_time']}")
Configuration
Environment Variables
The server uses default 12306 public APIs and doesn't require API keys. However, you can configure:
- Port (HTTP mode only): Set via
PORT environment variable (default: 8000)
- Timezone: Pass to
get_current_time tool (default: "Asia/Shanghai")
Station Data
The server includes a pre-built station database. To update:
cd scripts
uv run python update_station_data.py
This fetches the latest station list from 12306 and updates src/mcp_12306/data/station_names.json.
Project Structure
src/mcp_12306/
├── server.py # FastAPI main entry point
├── services/
│ ├── ticket_service.py # Ticket query logic
│ ├── station_service.py # Station search logic
│ └── http_service.py # 12306 API client
├── utils/
│ ├── config.py # Configuration
│ └── logger.py # Logging
├── data/
│ └── station_names.json # Station database
scripts/
├── start_server.py # HTTP mode launcher
└── update_station_data.py # Station data updater
Troubleshooting
"Station not found" errors
The station database uses exact matching. Use search_stations first to find the correct station name:
results = await client.call_tool("search_stations", {
"keyword": "beijing"
})
station_name = results[0]["name"]
No tickets available
- Check date format is YYYY-MM-DD
- 12306 typically allows booking 15-30 days in advance
- Use
get_current_time to get valid booking dates
- Try different dates or use
query_transfer for indirect routes
HTTP connection errors
- Ensure the server is running (
uv run python scripts/start_server.py)
- Check port 8000 is not in use
- For Docker, verify port mapping:
-p 8000:8000
- In China, direct 12306 API access may require proper network access
Stdio mode not working in Claude Desktop
- Restart Claude Desktop after configuration changes
- Verify JSON syntax in config file
- Check
uvx or pipx is installed: uvx --version
- For source mode, ensure
cwd points to correct directory
Rate limiting
The 12306 API may rate-limit requests. The server includes basic retry logic, but for production use:
- Add delays between requests
- Cache frequently accessed data (stations, routes)
- Use transfer queries sparingly (they make multiple API calls)
Best Practices
- Always get current time first: Use
get_current_time to determine valid booking dates
- Search stations before querying: Verify exact station names with
search_stations
- Check direct routes first: Try
query_tickets before query_transfer
- Cache station data: Station codes rarely change, cache
get_station_info results
- Handle no-results gracefully: Not all routes have direct or one-transfer options
- Use appropriate date formats: Always use YYYY-MM-DD for dates
Legal Notice
This project is for educational and research purposes only. It queries publicly available 12306 APIs and does not store, modify, or distribute official data. Users must comply with Chinese laws and 12306 terms of service. The authors assume no liability for misuse.