Did it work?For AI agents & developers

Developer changelog

What changed for integrators.

Changes to endpoints, tools, files and these docs. Subscribe: Atom feed of this page. Changes to published figures are in the donor changelog, What's changed (Atom).

  1. Quick answers to donor questions

    • Answer-first pages at /questions/ (for example /questions/best-charity-for-malaria): the cause's charities with an evaluator's top or promising designation, in the same order as the cause page, each with its headline figure and source. No ranking, scores or winners; like-for-like pairs link to /compare/.
    • Listed in site-index.json (questions: question, answer, charity_ids, claim_ids, comparisons, funds), llms.txt, llms-full.txt and the sitemap. The site-index schema documents the new array.

  2. Published figures in the open impact-claim format

    • Every modeled impact figure on a charity page is now also published as impact-claim/v0: all claims at /agents/claims.json, one file per charity at /charities/<slug>.claims.json (linked from each page with rel="alternate" and from site-index.json charities[].claims). Every file passes the reference validator in CI. See Schemas.
    • OG images and fonts are now served from Workers Static Assets instead of the Worker bundle. URLs, ETags and content types are unchanged; OG images now cache for a day.

  3. Side-by-side comparisons, ledger head, site-index schema and API catalog

    • Like-for-like comparison pages at /compare/: two charities rated by the same evaluator in the same unit and period, with both published figures and their sources. No ratios, differences or winners are computed. Listed in site-index.json (comparisons), llms.txt and llms-full.txt.
    • The outcome ledger head (record count and newest-record SHA-256) is published on /ledger/, in site-index.json (ledger_head) and in GET /v1/meta (ledger). New Atom feed of follow-up records: /ledger.xml.
    • New site-index JSON Schema, declared via $schema and enforced in CI, and an RFC 9727 /.well-known/api-catalog.
    • OpenAPI now has a public title (Did it work? API) and descriptions without internal build codes.

  4. Follow-up record pages, approaches index and a feed for this changelog

    • One permanent page per outcome-ledger record at /ledger/<record_id>, using the same id as GET /v1/ledger/{id}. Each shows the snapshot value, evaluator source, recorded and re-check dates, status, and the record and previous-record SHA-256, with schema.org Dataset JSON-LD.
    • /agents/site-index.json gains a ledger array (record id, page and API URLs, value, unit, dates, status, record_sha256); llms.txt and llms-full.txt list every record.
    • New approaches index at /interventions/ (was 404), grouped by the evaluators' own evidence ratings.
    • This changelog is now an Atom feed: /agents/changelog.xml.

  5. Developer section and MCP quickstart

    • All developer material now lives under /agents, with one sub-navigation on every page.
    • New: MCP quickstart with client configs and real captured responses, API reference generated from OpenAPI, schemas served at their $id URLs, data & feeds.
    • Technical HTML docs moved from /pages/… to /agents/reference/… (301 redirects). Raw /pages/*.md files are unchanged.

  6. Donor pages as data, feeds and embeds

    • One page per charity, cause and approach, generated from cited data, with schema.org JSON-LD.
    • New files: /agents/site-index.json, /llms-full.txt, /changes.xml, /sitemap.xml.
    • Embeddable evidence cards at /embed/<slug> with the /embed/v1.js loader (docs).
    • Anonymous usage counting on the site, API and MCP endpoint (no IPs, user agents, cookies or bodies).

  7. Outcome ledger and the impact-claim standard

    • GET /v1/ledger and GET /v1/ledger/{id}; MCP tools list_ledger_records and get_ledger_record.
    • Draft open format impact-claim v0 with a reference validator.
    • API and MCP endpoint public at https://api.diditwork.org.