worldgovdata API

There is one public base URL, and it is the only one. Everything below is relative to it.

The version lives in the path. /v1 will not change shape underneath you: fields are added, never removed or renamed. A breaking change would arrive as /v2.

The catalogue endpoints need no credentials at all, so the first call works before you have an account. This one asks the country spine for South Asia.

curl -sS "https://api.worldgovdata.com/v1/countries?region=South%20Asia"

{
  "data": [
    {
      "iso3": "AFG",
      "iso2": "AF",
      "m49": 4,
      "name": "Afghanistan",
      "official_name": "Islamic Republic of Afghanistan",
      "region": "South Asia",
      "subregion": "Southern Asia",
      "income_group": "Low income",
      "wb_code": "AFG",
      "capital": "Kabul",
      "wikidata_id": "Q889",
      "flag_emoji": "🇦🇫"
    },
    {
      "iso3": "BGD",
      "iso2": "BD",
      "m49": 50,
      "name": "Bangladesh",
      "official_name": "People's Republic of Bangladesh",
      "region": "South Asia",
      "subregion": "Southern Asia",
      "income_group": "Lower middle income",
      "wb_code": "BGD",
      "capital": "Dhaka",
      "wikidata_id": "Q902",
      "flag_emoji": "🇧🇩"
    }
    // … 6 more, ordered by name
  ],
  "meta": {
    "generated_at": "2026-07-28T09:14:02.481932+00:00",
    "count": 8,
    "region": "South Asia"
  }
}

Lines beginning // in the samples on this site mark where a real response was trimmed for reading. The API never returns comments.

To fetch observations rather than metadata you need a key. That is five minutes, and the quickstart walks the whole path — sign up, verify, create a key, make the first metered call, read the result.

Every JSON response is the same two keys. data is the payload — an array for list endpoints, an object for detail endpoints. meta describes the request that produced it and always carries generated_at.

{
  "data": [ /* rows, or a single object */ ],
  "meta": {
    "generated_at": "2026-07-28T09:14:02.481932+00:00",
    "count": 8
    // … the filters you sent, echoed back
  }
}

meta omits any field you did not send, so a key being absent means “not applied”, never “null”. Failures replace the whole body with a single error object:

{
  "error": {
    "code": "indicator_not_found",
    "message": "Unknown indicator: wdi.gdpp"
  }
}

Branch on error.code, never on error.message — the message is written for a human and may be reworded. Some codes add fields: a 402 carries required and balance, a 429 carries retry_after. The full list is on Errors.

Metadata is free and callable anonymously; the endpoints that return rows of observations cost credits. One credit is 100 rows.

A request that ends in an error costs nothing — including a too_many_rows rejection, which is counted before any data is read. New accounts get 2,000 credits at email verification. Full detail on Credits.

• Country — the spine. Keyed by ISO 3166-1 alpha-3 (IND, CHN), carrying region, subregion, income group, M49 and the World Bank code. Every observation hangs off one of these. Reference
• Indicator — one measured series, with an id like wgi.control_of_corruption: a provider prefix, a dot, a slug. Carries its unit, topic, direction (higher_is_better), coverage and, critically, the licence and attribution you must display. Reference
• Observation — one country, one indicator, one year, one value, sometimes a confidence interval. This is the atom, and /v1/data is how you get them. Reference
• Ranking — every country ordered on one indicator for one year, pre-computed, with the rank and the size of the ranked field. Reference

Topics group indicators into thirteen themes. Ask /v1/topics rather than hardcoding the list — it is data, and it grows.

• Read-only. Every route is a GET. Anything else returns 405 method_not_allowed.
• Ids are exact. Indicator ids are case-sensitive; ISO3 country codes are upper-cased for you.
• Cacheable — by you. An anonymous /v1 GET returns public, max-age=3600, s-maxage=21600. An authenticated one returns private, no-store, because it carries your balance and your remaining quota and must not sit in a shared cache. Cache it in your own application anyway: behind this data are pipelines that run daily at fastest and annually at slowest, and a response you did not re-request costs no credit.
• Compressed. Send Accept-Encoding: gzip and responses over roughly half a kilobyte come back gzipped. Every HTTP client does this by default.
• CORS is open, credentials are not. Any origin may call the API, but cookies are never accepted. Which is also why a key must never be in browser code — see Authentication.
• One hard cap. A single /v1/data request will not return more than 100,000 rows. Pagination has the recipe for fetching more than that.

The data is not ours. Each indicator arrives under its provider’s licence — most commonly CC BY — and every response tells you which: meta.licence and meta.attribution on the data endpoints, data.licence and data.attribution on the indicator record.