Labsco
MCP SERVER

Israeli Government Open Data

by DavidOsherProceed

Search Israel's data.gov.il catalogue in Hebrew or English, find the datastore behind a dataset, and query its rows with filters, sorting and pagination.

Government, Legal & Legislative RecordsVerified
Summary
Hebrew search that works before you know the dataset's name.

The bundled catalogue index is what separates this from a thin CKAN wrapper: normalized Hebrew, trigram fuzzy matching and tag co-occurrence mean a vague query in either language lands on real datasets, and the answer comes from memory rather than a round trip. From there it is a short path — resource, then rows.

What it is

An MCP server over the CKAN API behind data.gov.il, the Israeli government open-data catalogue. It ships a committed catalogue snapshot that is indexed in memory at startup, so discovery — search, tag ranking, organization listing — answers locally and falls back to live CKAN only when it has to. Tabular resources are then queried row by row. Every tool returns typed JSON in `structuredContent` and the same object as text for clients that read only text.

What you get
  • Dataset search over the local catalogue with live CKAN fallback, tolerant of Hebrew nikud, final-letter forms and typos — `find_datasets`
  • The full CKAN metadata for one dataset, and instant catalogue-backed summaries for all of them, optionally narrowed to one publisher — `get_dataset_info`, `list_all_datasets`
  • The files and datastores inside a dataset, flagged for which ones are query-ready — `list_resources`
  • Rows out of a datastore resource with full-text search, field filters, chosen columns, sorting, distinct values and pagination — `search_records`
  • The ministries, municipalities and agencies that publish, with Hebrew titles and dataset counts, plus live detail for one of them — `list_organizations`, `get_organization_info`
  • Topic tags ranked by real dataset count, with related-tag suggestions, and a keyword search across tag names — `list_available_tags`, `search_tags`
  • Six resources to read directly, including `datagov://featured` with ready-to-use resource ids and field schemas, and `datagov://catalog/stats` for what the snapshot contains
  • Three domain prompts with argument completions — food and nutrition, environment and sustainability, real-estate market
Requirements

No account and no key: `AUTH_MODE` defaults to `none` and data.gov.il is public. Local clients run `npx -y data-gov-il-mcp` (published as `data-gov-il-mcp` 3.0.1). For a shared instance, install globally and run the `data-gov-il-mcp-http` binary, or the container image, which serves Streamable HTTP on port 3664 with Helmet, CORS, rate limiting and Host/Origin validation in front. Exposing that instance is where the auth settings come in — `AUTH_MODE=apikey` with `API_KEYS`, or `AUTH_MODE=oauth` with issuer, audience and JWKS. 9 tools load by default; `summarize_dataset` appears only when `MCP_ENABLE_SAMPLING=true`, and the `interactive` clarification flow on search only when `MCP_ENABLE_ELICITATION=true`.

Setup effort

One command — npm install -g data-gov-il-mcp