StatCan Web Data Service MCP Server

A MCP (Model Context Protocol) server that provides access to Statistics Canada's Web Data Service, enabling AI assistants to discover, explore, analyze, and cite Canadian statistical data through natural language.

Project Overview

This server addresses several technical challenges in accessing Statistics Canada's Web Data Service (WDS) API:

1. API Format Requirements: Identified correct formats for StatCan WDS API endpoints to resolve 406 errors
2. Resilient Data Access: Implements multi-tier caching and fallbacks for API limitations
3. Enhanced Metadata: Provides rich context for statistical interpretation and proper citation
4. Analysis Capabilities: Includes statistical analysis, visualization, and forecasting features
5. MCP Integration: Connects with other MCP servers for expanded functionality

Features

• 🔍 Dataset Discovery: Search and browse StatCan datasets by keywords, themes, or geography
• 📊 Data Retrieval: Extract time series data with proper formatting for key vectors
• 📝 Metadata Exploration: Access detailed information about dataset structure and content
• 💾 Persistent Storage: Store datasets for future use with SQLite backend
• 📊 Advanced Analysis: Perform comprehensive statistical analysis, trend detection, seasonality analysis, and forecasting
• 📈 Visualizations: Generate data visualizations with integration to Vega-Lite
• 📑 Citations: Generate properly formatted citations for StatCan data
• 🖼️ Figure References: Track and reference figures created from StatCan data
• 🔄 API Resilience: Robust error handling with fallbacks for API limitations

Current Limitations

The StatCan WDS API has several limitations that this server addresses:

• Data Retrieval Constraints: Some API endpoints remain problematic despite correct formatting
• Format Sensitivity: Vector IDs must be numeric without the 'v' prefix, and payloads must be in array format
• Coordinate Access: Vector-based queries are more reliable than coordinate-based queries
• Performance Issues: Some API calls may timeout for large requests or during peak times
• Rate Limiting: High-volume queries may be throttled by the StatCan WDS API

The current implementation uses these strategies to work around these limitations:

• Multi-tier caching system at metadata, vector, and cube levels
• Local fallbacks for common statistical indicators
• Automatic format adjustment and retries with exponential backoff
• Graceful degradation to cached data when API endpoints fail

See docs/implementation_status.md and docs/api_connection_guide.md for details.

Quick Start

# Clone the repository
git clone https://github.com/yourusername/mcp-statcan.git
cd mcp-statcan

# Create a virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -e .

# Start the MCP server
python -m src

Dependencies

pip install sqlitedict aiohttp mcp pydantic python-dotenv pandas numpy

Usage with Claude

1. Open Claude Desktop App
2. Go to Settings > MCP Servers
3. Add a new server with the following configuration:
   • Name: StatCan Data
   • Command: path/to/venv/bin/python -m src
4. Start chatting with Claude and ask about Canadian statistics!

Working Example Queries

Here are queries that work reliably with the current implementation:

Basic Data Discovery

• "Find datasets about consumer prices in Canada"
• "What datasets do you have about employment?"
• "Show me the latest CPI data"

Vector-Based Data Retrieval

• "Get data for CPI vector 41690973"
• "Retrieve GDP data from vector 21581063"
• "Get the latest values for employment vector 111955426"

Analysis and Visualization

• "Generate a line chart for CPI data over the last 5 years"
• "Analyze the trend in GDP for the past 10 quarters"
• "Create a visualization of unemployment rate changes"

Citations

• "Generate a citation for the Consumer Price Index dataset"
• "How should I cite Statistics Canada's GDP data in APA format?"
• "Create a reference for the Labour Force Survey"

Testing

To verify the API client works correctly:

python -m tests.api.api_connection_steps

This runs step-by-step tests for:

• API connectivity
• Metadata retrieval
• Vector data access
• Format requirements

Project Structure

• /src - Core server implementation
• /docs - Documentation and guides
• /tests - Test suite for API and functionality
• /docs/references - API specifications and code sets

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgements

This project uses data from Statistics Canada, accessed via their Web Data Service API. It is not affiliated with or endorsed by Statistics Canada.

The Statistics Canada logo is used for informational purposes only to indicate the data source.