Starwell launches verified statistics API to stop AI agents from hallucinating numbers

Key takeaway

• Starwell launched a free API that gives AI agents access to real statistics from 29 official sources—the World Bank, Eurostat, Statistics Canada, FRED, and others—instead of letting them make up numbers.

• Every data point returned includes a citation to the official source, its license, and how it was retrieved, so AI systems can now answer questions about inflation, economic indicators, and labor data with verified facts rather than hallucinations.

3 Key Points

FAQ

What sources does Starwell connect to?

Starwell indexes 29 official sources including the World Bank Open Data (29,490 datasets), Eurostat (6,751 datasets), the International Labour Organization (1,958 datasets), Statistics Canada (8,226 datasets), the U.S. Federal Reserve's FRED API, the Australian Bureau of Statistics, and the U.S. Bureau of Economic Analysis, among others. The team adds more sources every week.

How do agents use Starwell?

Agents can POST to /v1/answer with a plain-English question (e.g., "compare inflation across four economies") and receive a computed answer that cites the official data points it used. Developers can also call the REST API directly (/sources, /datasets, /series, /observations) or integrate the MCP server into their agent stack.

Is there a cost?

Starwell offers a free tier with no credit card required, giving full access to the REST API, MCP server, computed answers, CSV export with citations, embeddable charts, webhook monitors, and the changelog feed.