pluraterra

Data for agents and developers

Public, read-only JSON and CSV access to captured country indicators and specialised evidence. Current release: 2026-11-02-r3. No API key or paid add-on is required.

Dataset catalogue · OpenAPI 3.1 · Agent discovery file · Release manifest and hashes

Agent reading guide and text alternatives · Page directory · New research feed

Start with a pinned query

curl 'https://www.pluraterra.com/api/v1/observations?country=PRT&indicator=life,air&from=2020&to=2025&release=2026-11-02-r3&limit=100'

Responses contain meta, data and links. Follow links.next until null. Pagination links pin the resolved release, including when the first request used “current”. This prevents a later release from changing a paginated retrieval midstream.

Query endpoints

EndpointDataset filters
/api/v1/observationscountry, indicator, provider, year, from, to, latest
/api/v1/householdscountry, scenario, year
/api/v1/regionscountry, geo, indicator, year
/api/v1/financial-straincountry, household, income_group, year
/api/v1/passport-routesorigin, destination, category
/api/v1/rankingscountry
/api/v1/robustnesscountry
/api/v1/countriescountry, region
/api/v1/indicatorsindicator, provider
/api/v1/sourcesid, area, q

All query endpoints accept release, limit (1–1000, default 100), offset (default 0), and format=json|csv. Country lists use comma-separated ISO3 codes. Duplicate and unsupported parameters are rejected. The catalogue accepts release and metadata search q.

Periods, dimensions and meaning

Observation rows carry a source code, unit, provider, original source URL, retrieval date, period start/end, observation type and dimensions. WHR rows are three-year survey means. OECD tax observations are explicit model households, not median residents. ILO hours are model estimates, not free time. Passport routes are conditional legal summaries with incomplete coverage, not a worldwide power ranking.

latest=true selects the latest endpoint within supplied filters per country and indicator. These endpoints can differ. Fixed-year queries do not fill missing years. Missing observations are omitted; absent legal routes are unreviewed/unknown. A null interval or status flag means it has not been ingested, not that error is zero.

CSV and caching

Add format=csv or send Accept: text/csv. Nested fields use JSON inside CSV cells. CSV pagination and total counts are exposed through Link, X-Total-Count and X-Pluraterra-Release headers. JSON has equivalent metadata in its envelope. ETags, conditional GET, HEAD and public CORS are supported. Use bulk immutable snapshots for large retrievals rather than many tiny requests.

Immutable downloads

Attribution and source terms

Preserve the returned release, dataset citation, original provider and source URL. The combined collection does not carry a blanket licence overriding provider rights. Eurostat reuse, World Bank dataset terms, OECD terms, ILO permissions and WHR data sharing remain relevant. Candidate sources in the catalogue are not automatically ingested or cleared for redistribution.

Errors and operational limits

400 means a malformed or unsupported query, 404 an unknown dataset/release, and 503 a captured snapshot could not load. Errors use application/problem+json. This is an initial service on one Basic dyno; no high-volume SLA is offered. Responses do not perform live provider fetches or calculate personal tax or immigration eligibility.

What makes an answer reproducible?

Record the exact release and query, follow all pages, retain the snapshot SHA-256, preserve units/dimensions and cite both Pluraterra and the original provider. Distinguish source estimates, normative index calculations, survey confidence intervals and weight-sensitivity ranges. Discovery files improve access; they do not guarantee that an AI system will cite the site.