GET /api/ask?q= · grounded, with citations
Ask the Almanac
Ask in plain language; get an answer composed strictly from the five Almanac datasets, with citations. No generative model sits in the loop — every sentence traces to a record. NLWeb in spirit: the site documents it, so it implements it.
try
how it works
Your question is tokenized, scored against every record in the Crawler Registry, Protocol Atlas,
Model Matrix and Lexicon, and the best matches are stitched into a plain-language answer with
grounded_in citations. It will only ever tell you what the Almanac actually contains —
if it doesn't know, it says so. That is the honest version of an answer engine.
$ curl -s "https://agentswelcome.dev/api/ask?q=who%20made%20x402"
{ "answer": "x402 (HTTP 402 Payments): Revives the dormant HTTP 402 status code…",
"grounded_in": [{ "type": "protocols", "id": "x402", "json": "/api/protocols/x402" }],
"confidence": "high" }