Data & feeds

Files you can fetch, cache and diff.

All regenerated from the repository data on every site build. Sizes are as of the last build.

FileFormatSizeWhat it holds
/llms.txtText98 KBDiscovery root for language models: what we are, how to use us, every donor page.
/llms-full.txtText42 KBFull text of the donor pages, with each figure, its evaluator, period and source.
/agents/site-index.jsonJSON49 KBEvery donor page with figures, units, periods, source URLs and embed links.
/agents/claims.jsonJSON48 KBEvery modeled impact figure on a charity page as open impact-claim/v0 claims (also one file per charity at /charities/<slug>.claims.json).
/changes.xmlAtom3 KBChanges to published figures, new pages and ledger records.
/sitemap.xmlXML10 KBEvery public page, with last-reviewed dates where known.
/robots.txtText843 BCrawl policy. Search engines and AI assistants are welcome.
/mcp-tools.jsonJSON15 KBMCP tool descriptors (the same tools the live endpoint lists).
/agents/guide.mdMarkdown32 KBThe agent guide, raw.
/schema/impact_claim.schema.jsonJSON Schema6 KBOpen impact-claim format, draft v0.
/schema/outcome_record.schema.jsonJSON Schema7 KBOutcome-ledger record, draft v0.
/ledger.xmlAtom6 KBOne entry per follow-up record (outcome ledger), newest first; re-check results will appear here.
/schema/site_index.schema.jsonJSON Schema13 KBFormat of /agents/site-index.json (v1), validated in CI.
/.well-known/api-catalogLinkset826 BRFC 9727 API catalog: OpenAPI, docs, meta and health links.

Live data from the API

  • GET https://api.diditwork.org/v1/claims: every numeric claim with evaluator, period and citations.
  • GET https://api.diditwork.org/v1/ledger: outcome-ledger records (what an evaluator said, when, and when to re-check).
  • GET https://api.diditwork.org/v1/leaderboards: one leaderboard per outcome unit.

Raw Markdown

Technical docs moved to /agents/reference/; the old /pages/… links redirect there. The raw files stay at their old addresses:

Caching

Every file is served with an ETag and a five-minute cache lifetime, so conditional requests (If-None-Match) are cheap.