canada-transit-mcpserver by heyyykk3 - MCP Server

Statistics Canada Transit MCP Server

48-Hour Challenge Submission - MCP for Statistics Canada LODE Dataset

🎯 What This Is

An MCP (Model Context Protocol) server that connects ChatGPT Desktop to Statistics Canada's Canadian Public Transit Network Database, providing access to GTFS (General Transit Feed Specification) data for 138 transit agencies across Canada.

📊 Dataset Information

Dataset: Canadian Public Transit Network Database
Source: Statistics Canada - Urban Data Lab
Release Date: January 31, 2025
Correction Date: May 7, 2025
URL: https://www150.statcan.gc.ca/n1/pub/23-26-0003/232600032025001-eng.htm
License: Open Government License - Canada
Total Agencies: 138 (covering all Canadian provinces and territories)

✨ Features

� 3 M CP Tools for dataset exploration and querying
📊 Complete GTFS Data Access - stops, routes, trips, schedules, shapes
🚏 138 Transit Agencies across Canada
🗺️ 50,000+ Transit Stops with coordinates
🚌 2,000+ Routes (bus, subway, rail, ferry)
🔗 HTTP/SSE Transport for ChatGPT Desktop
🔒 No Authentication Required (public data)
🐳 Docker Deployment on OMV/Linux servers
🌐 HTTPS Access via Tailscale Funnel

🚀 Quick Start

Prerequisites

Linux server (OMV/Debian) with Docker
Tailscale for HTTPS access
ChatGPT Desktop app

Installation

Step 1: Create project directory

ssh root@your-server
mkdir -p /opt/statcan-transit-mcp
cd /opt/statcan-transit-mcp

Step 2: Download the dataset

mkdir -p data
cd data
wget "https://www150.statcan.gc.ca/n1/en/pub/23-26-0003/2025001/zip/canadian_public_transit_network_database.zip?st=52TbFElb" -O canadian_public_transit_network_database.zip
apt install unzip -y
unzip canadian_public_transit_network_database.zip
cd ..

Step 3: Create server files

See TECHNICAL_REFERENCE.md for complete file contents:

statcan_transit_mcp/__init__.py
statcan_transit_mcp/data_loader.py
statcan_transit_mcp/http_server.py
pyproject.toml
Dockerfile
docker-compose.yml

Step 4: Build and run

docker-compose build --no-cache
docker-compose up -d

Step 5: Enable Tailscale Funnel

nohup tailscale funnel 3000 &

Your SSE URL: https://your-machine-name.your-tailnet.ts.net/sse

Step 6: Connect to ChatGPT Desktop

Open ChatGPT Desktop → Settings → MCP
Add new server
URL: https://your-machine-name.your-tailnet.ts.net/sse
Authentication: None
Save and enable

🔧 Available Tools (4 Tools)

1. `describe_dataset`

Browse/discover tool - Get complete dataset overview.

Example queries:

"Describe the Statistics Canada transit dataset"
"What file types are available?"
"How many different GTFS files exist?"

Returns:

Dataset name, source, license
Total agencies (138)
Total file types (39 different GTFS files)
List of all available file types
Core files that all agencies have
Usage instructions

2. `list_agencies`

List or search transit agencies across Canada.

Example queries:

"List all transit agencies"
"Find agencies in Montreal"
"Show me BC Transit agencies"
"Search for agencies in Ontario"

Returns:

Agency ID (folder name for use in queries)
Agency name
Website URL
Phone number
Number of available files for each agency

3. `get_agency_files` (NEW!)

Discover which GTFS files are available for a specific agency.

Example queries:

"What files does STM have?"
"Show me available data for TTC"
"List all files for Vancouver TransLink"

Parameters:

agency_id: Agency ID from list_agencies

Returns:

List of all .txt files available for that agency
File count
Core files vs optional files

Why this matters: Not all agencies have all 39 file types. Use this to discover what data exists before querying.

4. `query_data` (UNIVERSAL ACCESS)

Query data from ANY of the 39 GTFS file types for any agency.

Example queries:

"Get all stops from Montreal STM"
"Show me TTC routes"
"Get transfers data for Vancouver"
"Fetch fare_attributes for Ottawa"
"Get timetables for GO Transit"

Parameters:

agency_id: Agency folder name (from list_agencies)
file_name: Any GTFS file (e.g., 'stops', 'routes', 'transfers.txt', 'fare_attributes')
limit: Max records (default 5000, max 100000)

Supports ALL 39 File Types:

Core (all 138 agencies): agency, routes, stops, stop_times, trips
Near-universal (95%+): shapes, calendar_dates
Common (50%+): feed_info, calendar
Optional (10-30%): fare_attributes, transfers, fare_rules, frequencies
Specialized (1-5 agencies): timetables, pathways, levels, attributions, and 25 more

Returns:

Complete file data as JSON
Record count
Attribution
Available files if requested file doesn't exist

📁 Project Structure

/opt/statcan-transit-mcp/
├── statcan_transit_mcp/
│   ├── __init__.py
│   ├── data_loader.py        # GTFS data access
│   └── http_server.py         # MCP server (3 tools)
├── data/
│   └── canadian_public_transit_network_database/
│       ├── gtfs/              # 138 agency folders
│       ├── data_sources.csv   # Agency metadata
│       └── stops_and_routes.gpkg  # Geospatial data
├── docker-compose.yml
├── Dockerfile
└── pyproject.toml

🎮 Usage Examples

In ChatGPT Desktop:

Dataset Discovery

"Describe the Statistics Canada transit dataset"
→ Returns: 138 agencies, provinces, data types, schemas

Agency Search

"Find transit agencies in Montreal"
→ Returns: STM (Société de transport de Montréal) with ID and details

"List all BC Transit agencies"
→ Returns: 30+ BC Transit regional systems

Data Queries

"Get 10 stops from Montreal STM"
→ Returns: Stop names, coordinates, codes

"Show me all TTC subway routes"
→ Returns: Line 1 (Yonge-University), Line 2 (Bloor-Danforth), etc.

"Get stop times for route 161 in STM"
→ Returns: Arrival/departure times for each stop

Complex Analysis

"Get all stops from STM and find ones with 'Patricia' in the name"
→ ChatGPT gets all stops, filters for 'Patricia'
→ Returns: Sherbrooke / Patricia, Poirier / Patricia with coordinates

🔍 What Data Is Available - 39 GTFS File Types!

The dataset contains 39 different GTFS file types across 138 agencies. Here's the complete breakdown:

Core Files (100% - All 138 Agencies Have These)

✅ agency.txt (138) - Agency contact info, timezone, phone
✅ routes.txt (138) - All routes with names, types, colors
✅ stops.txt (138) - All transit stops with coordinates
✅ stop_times.txt (138) - Arrival/departure times
✅ trips.txt (138) - Individual trip schedules

Near-Universal (95%+)

✅ shapes.txt (136) - Geographic route paths for mapping
✅ calendar_dates.txt (134) - Service exceptions (holidays)

Common (50%+)

✅ feed_info.txt (106) - Dataset metadata and version
✅ calendar.txt (74) - Regular service schedules

Optional (10-30%)

fare_attributes.txt (39) - Fare pricing information
transfers.txt (33) - Transfer rules between routes
fare_rules.txt (31) - Fare application rules
frequencies.txt (14) - Headway-based service

Specialized (1-5 agencies)

directions.txt (5) - Direction names
timetables.txt (4) - Timetable definitions
timetable_stop_order.txt (3) - Stop order in timetables
stop_attributes.txt (3) - Additional stop metadata
runcut.txt (3) - Run cutting information
location_groups.txt (3) - Location groupings
linked_datasets.txt (3) - Dataset relationships
farezone_attributes.txt (3) - Fare zone details
calendar_attributes.txt (3) - Calendar metadata
booking_rules.txt (3) - Booking requirements
areas.txt (3) - Geographic areas
stop_amentities.txt (2) - Stop amenities
__Licence.txt (2) - License information

Rare (1 agency each)

27-39. trip_pattern.txt, translations.txt, timetable_pages.txt, timetable_notes.txt, timetable_notes_references.txt, stop_order_exceptions.txt, signup_periods.txt, route_names_exceptions.txt, route_directions.txt, pathways.txt, levels.txt, direction_names_exceptions.txt, attributions.txt

Total: 39 different file types!

Use get_agency_files tool to see which files each agency has, then query_data to access any file.

🛠️ Management Commands

# View logs
docker-compose logs -f

# Restart server
docker-compose restart

# Stop server
docker-compose stop

# Check health
curl http://localhost:3000/health

# Test tools list
curl -X POST http://localhost:3000/ -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

# Rebuild after changes
docker-compose down
docker-compose build --no-cache
docker-compose up -d

🌐 Access URLs

Local: http://localhost:3000
Tailscale Funnel: https://your-machine-name.your-tailnet.ts.net
Health check: /health
SSE endpoint: /sse (for ChatGPT connection)
JSON-RPC: / (POST)

📝 Real-World Example

You ask ChatGPT: "Find stops with 'patricia' in Montreal STM"

ChatGPT workflow:

Uses list_agencies with query="montreal" → gets agency_id: "societe_transport_montreal"
Uses query_data with agency_id="societe_transport_montreal", data_type="stops", limit=10000
Filters results for stops containing "patricia"
Returns:
- Sherbrooke / Patricia (stop 50747) - 45.455402, -73.641252
- Poirier / Patricia (stop 55917) - 45.519952, -73.696873

� Troubleshooting

Container won't start:

docker-compose logs

Port already in use:

docker ps
# Stop conflicting container or change port in docker-compose.yml

Tailscale Funnel not working:

tailscale status
# Restart: nohup tailscale funnel 3000 &

ChatGPT can't connect:

Verify Tailscale Funnel is running
Check URL format: https://machine-name.tailnet.ts.net/sse
Ensure server is enabled in ChatGPT MCP settings
Try disconnecting and reconnecting

Tools not showing in ChatGPT:

Rebuild container: docker-compose down && docker-compose build --no-cache && docker-compose up -d
Completely restart ChatGPT Desktop
Verify tools: curl -X POST http://localhost:3000/ -H "Content-Type: application/json" -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

�� Dataset Statistics

Total Agencies: 138 (confirmed from data_sources.csv)
Total File Types: 39 different GTFS files
Total Stops: ~50,000+ across all agencies
Total Routes: ~2,000+ across all agencies
Provinces Covered: All Canadian provinces and territories
Data Format: GTFS (CSV .txt files)
Geospatial Format: GeoPackage (.gpkg)
Dataset Size: 424 MB (compressed zip)
File Coverage:
- 5 core files: 100% (all 138 agencies)
- 2 near-universal: 95%+ coverage
- 4 common: 50%+ coverage
- 28 specialized: 1-30% coverage

🎯 Challenge Requirements Met

✅ MCP Server: HTTP/SSE transport compatible with ChatGPT
✅ No Authentication: Public endpoint, no credentials required
✅ Browse/Discover Tool: describe_dataset provides complete schema and metadata
✅ Query Tools: 4 tools (list_agencies, get_agency_files, query_data) return structured JSON
✅ Dataset Source: Statistics Canada LODE - Canadian Public Transit Network Database (138 agencies)
✅ Schema Introspection: get_agency_files shows available data, describe_dataset lists all 39 file types
✅ Universal Data Access: Supports ALL 39 GTFS file types automatically
✅ Error Handling: Helpful error messages with suggestions and available files list
✅ License Attribution: Open Government License - Canada properly attributed in all responses
✅ Performance: Queries return within 5-10 seconds
✅ No Tracking: No analytics, telemetry, or PII collection
✅ Reliability: Docker restart policy, graceful error handling

📄 License & Attribution

Server Code: Open source
Dataset: Open Government License - Canada
GTFS Specification: Creative Commons Attribution 3.0

Required Citation: Statistics Canada. (2025). Canadian Public Transit Network Database. https://www150.statcan.gc.ca/n1/pub/23-26-0003/232600032025001-eng.htm

🤝 Credits

Dataset: Statistics Canada - Urban Data Lab, Centre for Special Business Projects
Supported by: Housing, Infrastructure and Communities Canada (HICC)
Data Contributors: 138 Canadian transit agencies who provide open GTFS data

📧 Support

For dataset issues: statcan.lode-ecdo.statcan@statcan.gc.ca
For MCP server issues: Check logs and TECHNICAL_REFERENCE.md
For ChatGPT integration: Refer to OpenAI's MCP documentation

Version: 1.0.0
Release Date: May 2025
Dataset Release: January 31, 2025 (Corrected: May 7, 2025)
Built for: 48-Hour MCP Challenge
Agencies: 138 across Canada
File Types: 39 different GTFS files
Tools: 4 (describe_dataset, list_agencies, get_agency_files, query_data)
Data Access: Universal - ALL file types supported

heyyykk3/canada-transit-mcpserver

Statistics Canada Transit MCP Server

🎯 What This Is

📊 Dataset Information

✨ Features

🚀 Quick Start

Prerequisites

Installation

🔧 Available Tools (4 Tools)

1. describe_dataset

2. list_agencies

3. get_agency_files (NEW!)

4. query_data (UNIVERSAL ACCESS)

📁 Project Structure

🎮 Usage Examples

Dataset Discovery

Agency Search

Data Queries

Complex Analysis

🔍 What Data Is Available - 39 GTFS File Types!

Core Files (100% - All 138 Agencies Have These)

Near-Universal (95%+)

Common (50%+)

Optional (10-30%)

Specialized (1-5 agencies)

Rare (1 agency each)

🛠️ Management Commands

🌐 Access URLs

📝 Real-World Example

� Troubleshooting

�� Dataset Statistics

🎯 Challenge Requirements Met

📄 License & Attribution

🤝 Credits

📧 Support

1. `describe_dataset`

2. `list_agencies`

3. `get_agency_files` (NEW!)

4. `query_data` (UNIVERSAL ACCESS)