Train Ticket Query MCP Server (FastMCP)

A train/high-speed rail query service implemented with FastMCP, which can be called as a remote MCP tool by any Agent (such as deepagents, LangGraph, Claude, etc.).

Capabilities

One tool function: query_train_tickets(from_station, to_station, date)

• Query the train numbers, schedules, durations, booking availability, and remaining tickets between two stations on a given day

• Data source: 12306 public query interface (real data); automatically switches to simulated data when the interface is blocked, with the source clearly marked, ensuring the pipeline can be demonstrated at any time

Related MCP server: YHT Bilet MCP Server

Local Run

pip install -r requirements.txt
python server.py        # 启动 streamable-http 服务，默认 http://127.0.0.1:8000

MCP client configuration (local debugging):

{
  "mcpServers": {
    "train": {
      "url": "http://127.0.0.1:8000/mcp/",
      "transport": "streamable_http"
    }
  }
}

Deploy to ModelScope MCP Plaza (Key Steps)

1. Push this repository to GitHub

2. Open modelscope.cn/mcp → Log in → Click "Create MCP"

3. Select "GitHub Quick Create" as the creation type; select "Hosted Deployment" as the hosting type

4. Fill in the English/Chinese name, source URL (GitHub repository URL), and complete the configuration as prompted by the platform

5. After submission, the platform generates an access URL: https://mcp.api-inference.modelscope.net/<your ID>/sse or /mcp (can be set to be valid indefinitely)

6. Fill this URL back into the MCP configuration of the Agent project

Tip: The platform has startup method requirements for hosted projects (usually requires Python + starting an HTTP service). If deployment fails, adjust the entry point as prompted by the platform (e.g., provide a run script or environment variable).

Data Source Description

• 12306 has no official open API. This project uses its public query interface (kyfw.12306.cn/otn/leftTicket) for querying without login, for personal learning and research only

• The interface has anti-bot measures (occasional 302 redirects). The code includes session cookies + simulated data as a fallback

• If more stable data is needed, you can replace _fetch_12306 with a third-party data source (pluggable)

Directory

• server.py — FastMCP service main program

• test_client.py — Local debugging script (directly calls tool functions + tests via MCP protocol)

• requirements.txt — Dependencies