datagov-mcp

An MCP server for exploring Israeli government open data (data.gov.il) — with built-in interactive visualizations powered by MCP Apps.

Search thousands of public datasets, profile their structure, generate charts, and plot geographic data on maps — all from your AI assistant.

Profile public housing datasets to understand unit sizes, locations, and availability:

See which cities have the most demand in government housing lotteries:

Understand the distribution of apartment sizes across the country:

Track housing unit availability over time:

Visualize education institutions across Israel:

Plot public transport stations:

git clone https://github.com/aviveldan/datagov-mcp.git
cd datagov-mcp

# Create virtual environment and install (requires uv)
uv venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate
uv pip install -e ".[dev]"

fastmcp install claude-desktop server.py

Restart Claude Desktop — you'll see the DataGovIL tools available immediately.

fastmcp dev inspector server.py

This opens a web UI where you can browse tools, test them interactively, and preview MCP App visualizations in the Apps tab.

Preview the interactive visualization apps locally:

fastmcp dev apps server.py

Here's a real workflow for someone exploring the Israeli housing market:

You: "Search for discounted housing lottery datasets"

→ package_search(q="דירה בהנחה")
  Found: "נתונים תקופתיים - תכנית דירה בהנחה" (Discounted Housing Program)
  Resource ID: 7c8255d0-49ef-49db-8904-4cf917586031

You: "Profile this dataset so I can understand what fields are available"

→ dataset_profile(resource_id="7c8255d0-49ef-49db-8904-4cf917586031")
  Shows: LamasName (city), Subscribers, Winners, PriceForMeter,
         LotteryHousingUnits, ProjectName, Neighborhood...

You: "Show me which cities have the most subscribers competing for units"

→ chart_generator(
    resource_id="7c8255d0-49ef-49db-8904-4cf917586031",
    chart_type="bar",
    x_field="LamasName",
    y_field="Subscribers",
    title="Housing Lottery Subscribers by City"
  )
  → Interactive bar chart rendered in MCP Apps UI

You: "Now show the public housing units — map them and show me sizes"

→ dataset_profile(resource_id="c3a68837-9b7a-4ee7-bd92-130678dc8ae3")
  Shows: CityLmsName, NumOfRooms (avg 2.4), Floor, TotalArea (21-110 m²)...

→ chart_generator(
    resource_id="c3a68837-9b7a-4ee7-bd92-130678dc8ae3",
    chart_type="histogram",
    x_field="TotalArea",
    title="Distribution of Housing Unit Sizes (m²)"
  )
  → Most units are 48-57 m², with a long tail up to 110 m²

Insight: Cities like Ashkelon and Sderot show 25,000-35,000 subscribers per lottery — that's intense competition. Smaller cities in the periphery (Umm al-Fahm, Nazareth) have far fewer. If you're flexible on location, your odds improve dramatically.

These tools render interactive UI directly in MCP-compatible clients.

Profile a dataset to understand its structure and quality.

• Fields detected: integer, number, string, coordinate
• Statistics: min, max, mean, null count, unique values
• Output: Interactive DataTable with search/filter

dataset_profile(resource_id="c3a68837-9b7a-4ee7-bd92-130678dc8ae3", sample_size=200)

Generate interactive charts from any dataset.

chart_generator(
  resource_id="7c8255d0-49ef-49db-8904-4cf917586031",
  chart_type="bar",
  x_field="LamasName",
  y_field="Subscribers",
  title="Housing Lottery Subscribers by City",
  limit=50
)

Plot geographic data on interactive Leaflet maps.

map_generator(
  resource_id="e873e6a2-66c1-494f-a677-f5e77348edb0",
  lat_field="Lat",
  lon_field="Long",
  limit=500
)

Here are some interesting datasets to get started with:

Visualization tools use FastMCPApp providers with prefab-ui components:

• DataProfile app → DataTable, Metric components
• Charts app → BarChart, LineChart, ScatterChart, Histogram
• Maps app → Embed with Leaflet HTML

Tools registered via @app.ui() automatically get proper MCP Apps metadata and render in compatible clients.

All API calls use httpx.AsyncClient with:

• 30-second timeout
• Automatic retries for 5xx errors
• Connection pooling

• Numeric values from CKAN are coerced (handles "25" → 25.0)
• Map popup content is HTML-escaped to prevent XSS
• Line charts are sorted by x-axis for correct rendering

pytest tests/ -v          # 39 tests
pytest tests/ --cov=datagov_mcp  # With coverage

ruff check .   # Lint
ruff format .  # Format

datagov-mcp/
├── datagov_mcp/
│   ├── server.py          # Core CKAN tools + provider registration
│   ├── apps.py            # FastMCPApp definitions (DataProfile, Charts, Maps)
│   ├── visualization.py   # Visualization tools (@app.ui entry points)
│   ├── api.py             # CKAN API helper
│   └── client.py          # HTTP client
├── tests/                 # 39 tests with HTTP mocking
│   ├── test_api.py
│   ├── test_contracts.py
│   ├── test_tools.py
│   └── test_visualization.py
├── screenshots/           # Auto-generated demo screenshots
├── server.py              # Entrypoint
└── pyproject.toml

We welcome contributions! See CONTRIBUTING.md for guidelines.

1. Fork the repository
2. Create a feature branch
3. Make changes with tests
4. Submit a pull request

pip install nano-dev-utils
python -c "from nano_dev_utils import release_ports; release_ports.PortsRelease().release_all()"

Avoid running installation in OneDrive-synced folders. See uv#7906.

uv pip install -e ".[dev]"

MIT — see LICENSE.

• Built with FastMCP and prefab-ui
• Data from Israel Open Data Portal
• Maps powered by Leaflet + CARTO
• Charts powered by Recharts (via prefab-ui)