zurich-opendata

This MCP server exposes Zurich Open Data through a set of 20 tools and 6 resources, enabling AI assistants like Claude or ChatGPT to query datasets, geospatial data, real-time environmental and mobility information, tourism data, and SPARQL-backed linked data. Core capabilities include CKAN Open Data search and retrieval (e.g., zurich_search_datasets, zurich_get_dataset), real-time environmental feeds (zurich_weather_live, zurich_air_quality, zurich_water_weather), mobility stats (zurich_pedestrian_traffic, zurich_vbz_passengers, zurich_parking_live), geospatial access (zurich_geo_layers, zurich_geo_features), Gemeinderat data (zurich_parliament_search, zurich_parliament_members), tourism and statistics (zurich_tourism, zurich_sparql), and data analysis tools (zurich_analyze_datasets, zurich_catalog_stats, zurich_find_school_data). You can query these tools via natural language prompts and have the MCP server translate requests into the appropriate API calls and data formats. Use the configured MCP server name zurich-opendata in your client (Claude, VS Code, or CLI) to route questions to the corresponding tool via the MCP protocol.

Prerequisites:

• Python 3.11+
• pip or uvw

Installation steps:

1. Clone the repository:

git clone https://github.com/schulamt-zuerich/zurich-opendata-mcp.git
cd zurich-opendata-mcp

2. Install in editable mode:

pip install -e .

3. (Optional for uv) If using uv, install with uv in front of the command as shown:

udb

4. Run the MCP server (examples):

python -m zurich_opendata_mcp.server

5. Verify the server starts and is reachable by your MCP client (Claude, VS Code integration, or CLI) using the server name zurich-opendata.

Note: The project provides alternative startup methods in the README, such as using a ready-made CLI entry if installed as an executable, but the standard approach is to run the Python module as shown above.

Tips and common issues:

• Ensure Python 3.11+ is used to match the package requirements.
• If you encounter dependency issues, install with the editable flag (pip install -e .) to ensure local changes are reflected.
• The MCP configuration uses the server name zurich-opendata. If you run Claude or VS Code, ensure their mcpServers entry matches this name.
• Environment variables can be populated in the mcpServers["zurich-opendata"].env dictionary if your deployment requires API keys or custom endpoints.
• When testing, start the server locally and verify endpoints with a simple MCP client or curl-based test harness to confirm tool availability before full integration.