SQL executes inside DuckDB-Wasm in the visitor's browser. It does not consume Worker CPU.
Find Belgian
open data you can trust
Find the right dataset, understand what it contains, and see the evidence behind its freshness, size and usability before you rely on it.
0source systems
0guided workflows
0identifier concepts
0typed relationships
⌕/
START WITH
SQL READY compatible distribution and direct browser access verifiedSQL POSSIBLE compatible distribution; browser access not yet verifiedDOWNLOAD ONLY browser access failed or a conversion/download step is neededNOT TABULAR no directly queryable table-like distributionPreviewable a small sample can be inspected in-browser; this does not guarantee SQL loadingOther downloads PDFs, archives and unsupported files are kept separate and can be filtered outCurated manually assembled from official publisher material; evidence links includedPublisher schema machine-harvested field definitions exposed by the publisherNational DCAT discovered through data.gov.be; direct publisher evidence wins when both existHarvested catalogue metadata; grain not auditedRecord sample actual example rows, never treated as schema
NO MATCH
No source matches this combination.
Source-level graph with explicit relationship contracts. An edge can be dataset-scoped, lossy or contextual; connectivity never means universal record-level joinability.
Runs in your browser. Load files, query them with DuckDB SQL, test join cardinality and inspect duplicate keys before you trust a merge. Local files are not uploaded.
01 / INGEST
Load data
engine idle
Remote URLs are tried directly first. For catalogue CSV/TSV files blocked by CORS, SQL Lab may fetch the complete file through the approved opaque-ID relay when it fits within 8 MiB. Arbitrary URLs are never relayed.
No tables loaded yet.
02 / QUERY
Ctrl / ⌘ + Enter to runWrite a SELECT
No resultDisplay is capped at 500 rows.
Run a SELECT to see rows here.
03 / VISUALIZE
Visual Data Studio
Turn the current SQL query into charts and pivots, right in your browser. Use the demo to explore the controls without loading publisher data.
Run a query or try the visual demo.
The visualization uses at most 10,000 query rows and warns when capped. Exported view settings contain no dataset. The Perspective engine loads from a third-party CDN only when used.
04 / JOIN CHECK
Check cardinality before mergingBuild & test a join
Load two tables to test match rate, null keys, duplicates and row multiplication.
Duplicate explorer
Show every row whose selected key occurs more than once.
05 / PRACTICAL LIMITS
What stays free
Files stay local to the browser session. No upload endpoint is involved.
The browser contacts publishers directly first. Catalogue CSV/TSV blocked by CORS can use the complete-file ≤8 MiB relay; larger files stay download/drop only unless direct CORS works.
DuckDB-Wasm is loaded from jsDelivr on first use, keeping large WASM binaries out of this GitHub repo and Cloudflare assets.
This compares publishers and access routes; use the Dataset Atlas to inspect individual fields, versions and record-level compatibility.