Describe a domain in one sentence — “Swedish VC firms and business angels”, “tantric events in Sweden”, “Japanese crafts and courses in Stockholm”. Nolvin's agents find the sources, crawl them politely, extract the people, organizations, places and events with a quote for every fact, and keep the graph current. You query it through one API.
From one sentence, an agent writes a brief — scope, inclusion rules, expected examples, search queries — then discovers and validates sources. Crawling honors robots.txt and spaces requests per host. How it works →
Pages become markdown, then typed entities and relationships — each with the page and verbatim snippet it came from. Re-reads are incremental: an unchanged page costs nothing, a changed one is read as a diff. The POLE model →
Person, Object, Location, Event. Graph traversal, timeline and map — every kind projects through its relationships. REST, webhooks and share links on top. Read the docs →
$ curl https://api.nolvin.com/v1/domains/autopilot \
-H "Authorization: Bearer $NOLVIN_KEY" \
-d '{"prompt":"Tech consulting firms in Linköping and Norrköping"}'
{"id":"6ba1539b-…","stage":"brief"}
$ curl https://api.nolvin.com/v1/domains/6ba1539b-…/entities?type=object
{"entities":[{"name":"HiQ","type":"object", …}], "next_offset":50}
Brief, discovery, crawl, extraction, assessment and steady-state upkeep — the autopilot loop end to end.
read → 02 / modelPerson, Object, Location, Event: why four kinds, their facets, canonical relations and projections.
read → 03 / examplesEvent calendars, local industry maps, investor networks and cultural scenes built from one sentence.
read → 04 / developersQuickstart, authentication, graph queries, webhooks and the full endpoint reference with OpenAPI.
read →