🇨🇭 Part of the Swiss Public Data MCP Portfolio

🚆 swiss-transport-mcp

MCP server connecting AI models to the Swiss public transport system – journey planning, real-time departures, disruptions, occupancy, ticket prices, train formations and open data from opentransportdata.swiss.

🇩🇪 Deutsche Version

Demo

Overview

swiss-transport-mcp gives AI assistants like Claude a complete Swiss travel information system – not just timetables, but also real-time disruption alerts, occupancy forecasts, ticket prices, and a full train formation view. All accessible through a single, standardised MCP interface.

The various APIs at opentransportdata.swiss speak different protocols – OJP 2.0 (XML/SOAP), SIRI-SX (XML), REST/JSON. This server translates everything into clean JSON for the AI model, acting as a multilingual protocol interpreter.

Anchor demo query: "Plan a school trip for 25 students from Zurich to the Technorama in Winterthur – check for disruptions and find the best departure." → More use cases by audience →

Features

• 🗺️ Journey planning (A → B with transfers, duration, transport mode) via OJP 2.0
• 🕐 Real-time departures with delays and platform information
• 🔍 Stop search by name or coordinates
• 🚨 Live disruption alerts (cancellations, closures) via SIRI-SX
• 📊 Occupancy forecasts for trains (SBB, BLS, Thurbo, SOB)
• 💰 Ticket prices including class selection
• 🚃 Train formation – coaches, classes, amenities, accessibility
• 📦 Open data catalogue – ~90 transport datasets via CKAN
• 🔑 Graceful degradation – server starts with core tools even without optional API keys
• ☁️ Dual transport – stdio for Claude Desktop, Streamable HTTP/SSE for cloud deployment

Prerequisites

• Python 3.11+
• A free API key from api-manager.opentransportdata.swiss (subscribe to OJP 2.0 as minimum)
• Optional: additional keys for SIRI-SX, Occupancy, Formation, OJP Fare

Installation

# Clone the repository
git clone https://github.com/malkreide/swiss-transport-mcp.git
cd swiss-transport-mcp

# Install
pip install -e .

Or with uvx (no permanent installation):

uvx swiss-transport-mcp

Quickstart

# Set the minimum required key (OJP core tools)
export TRANSPORT_API_KEY=your_key_here

# Start the server (stdio mode for Claude Desktop)
swiss-transport-mcp

Try it immediately in Claude Desktop:

"What are the next departures from Zurich Stadelhofen?" "How do I get from Wädenswil to Bern by train?"

Configuration

Environment Variables

APIs without a key are silently disabled – the server starts fine with just the 6 core tools.

Operational / security variables:

🔒 Egress allow-list: all outbound requests are restricted to https:// on opentransportdata.swiss hosts. Any other host is refused before a request is sent (SSRF / egress hardening).

Claude Desktop Configuration

Minimal (core tools only):

{
  "mcpServers": {
    "swiss-transport": {
      "command": "swiss-transport-mcp",
      "env": {
        "TRANSPORT_API_KEY": "your_key_here"
      }
    }
  }
}

Full (all 11 tools):

{
  "mcpServers": {
    "swiss-transport": {
      "command": "swiss-transport-mcp",
      "env": {
        "TRANSPORT_API_KEY": "your_ojp_key_here",
        "SIRI_SX_API_KEY": "your_siri_key_here",
        "OCCUPANCY_API_KEY": "your_occupancy_key_here",
        "FORMATION_API_KEY": "your_formation_key_here",
        "OJP_FARE_API_KEY": "your_fare_key_here"
      }
    }
  }
}

Config file locations:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

Cloud Deployment (Streamable HTTP)

For use via claude.ai in the browser (e.g. on managed workstations without local software). The cloud transport is Streamable HTTP (MCP_TRANSPORT=streamable-http, endpoint /mcp). SSE (/sse) is still supported but deprecated.

Docker (recommended):

# Build + run with explicit resource limits (see docker-compose.yml)
TRANSPORT_API_KEY=xxx docker compose up --build
# → http://127.0.0.1:8000/mcp

The image is a multi-stage build running as a non-root user; docker-compose.yml adds read_only, no-new-privileges and memory/CPU/PID limits.

Render.com:

1. Push/fork the repository to GitHub
2. On render.com: New Web Service → connect GitHub repo (Docker runtime)
3. Set env MCP_TRANSPORT=streamable-http and MCP_HOST=0.0.0.0
4. In claude.ai under Settings → MCP Servers, add: https://your-app.onrender.com/mcp

💡 "stdio for the developer laptop, Streamable HTTP for the cloud."

Scaling horizontally: run with MCP_STATELESS=1. In stateless mode the server keeps no per-session state, so any instance can serve any request and a plain round-robin load balancer suffices — no sticky sessions / Mcp-Session-Id affinity required. If you need stateful streaming instead, route by Mcp-Session-Id at the edge LB (e.g. HAProxy stick-tables) so each session stays pinned to one instance.

⚠️ Binding: In a network transport the server binds to 127.0.0.1 by default so a locally started server is not exposed to your whole network (e.g. public Wi-Fi). Set MCP_HOST=0.0.0.0 only in a container/cloud environment where binding to all interfaces is intended (the Docker image does this for you).

Available Tools

Core Tools (OJP 2.0 / CKAN)

¹ CKAN tools require a separate subscription in the API Manager.

Extension Tools (optional API keys)