Plug the whole dataset into anything
Static JSON, no key, no rate limit, CORS-open by virtue of being files on a CDN. Designed so an AI assistant can answer benefits questions from it directly.
REST-shaped endpoints
/api/v1/countries.json | Country index: codes, currencies, regions, income bands, counts. |
|---|---|
/api/v1/programmes/{cc}.json | Every programme for one country, full records. |
/api/v1/stats.json | Live dataset statistics — the same numbers this site renders. |
/api/v1/mcp-tools.json | Tool schemas (JSON Schema 2020-12) for the MCP layer. |
/llms.txt | Plain-text orientation for language models. |
MCP layer
The MCP tool definitions ship as data at /api/v1/mcp-tools.json. Any MCP server that can
fetch JSON can expose these tools with a thin adapter — the tools are pure reads over the static files,
so the server needs no database and no state:
list_countries | Coverage and counts. |
|---|---|
search_programmes | Filter by country, category, keyword, verification status. |
get_programme | Full record including steps, documents and source. |
match_profile | Run the eligibility engine over a profile and return the three buckets. |
Status: schemas shipped, server not deployed. We won't call a documented
interface a running one. The matcher itself (src/engine/matcher.js) is dependency-free and
imports into a Node MCP server unchanged.
Licence
MIT. Attribution appreciated, not required. The underlying programme information belongs to the publishing authorities and is linked on every record.
Every number on this site is generated from these files at build time. If the API and a page ever disagree, the API is right and it is a bug — please report it.