Stop scraping.
Start querying.

Sources

Ingested from sources you already trust

• Bank of Canada Valet 8 datasets
• European Central Bank 4 datasets
• Eurostat 4 datasets
• FRED (St. Louis Fed) 10 datasets
• Government of Canada Open Data 10 datasets
• ILO ILOSTAT 4 datasets
• IMF DataMapper 6 datasets
• NASA POWER 2 datasets
• OECD 4 datasets
• Office for National Statistics 4 datasets
• Our World in Data 2 datasets
• WHO Global Health Observatory 8 datasets
• World Bank 10 datasets

How it works

Raw source to clean table in three steps

1. 01

   Connectors ingest

   One connector per source fetches raw data on a schedule and maps it to a canonical schema — dates, units, and country codes normalized.

   world_bank · fred · who · imf
2. 02

   Snapshots freeze

   Every refresh becomes an immutable Parquet snapshot with a content hash, quality score, and changelog — history you can cite.

   sha256 · parquet · versioned
3. 03

   You query or download

   SQL in the workbench, one-click CSV/JSON/Parquet, a REST API for apps, and an MCP server for AI agents.

   sql · rest · mcp · csv

Access

One dataset, every interface

SELECT year, value
FROM ds_world_bank_inflation_cpi
WHERE country_code = 'CAN'
ORDER BY year DESC LIMIT 5;

GET /v1/datasets/world_bank/gdp_current_usd/data?format=parquet
X-API-Key: $HUB_KEY

{
  "mcpServers": {
    "hub": {
      "command": "python",
      "args": ["-m", "hub.mcp_server"]
    }
  }
}