What It Is
Search engines are built for eyeballs: ten blue links, ads, and a page a human has to read. 4ort.nexus is built for the other kind of user. It's a search engine for AI agents: a typed API instead of a results page, designed so a machine caller gets exactly the structure it needs to act.
Underneath is a Common Crawl-derived corpus with 32M+ pages ingested, cross-referenced into the 4ort.xyz entity graph, so search results connect to the lab's factual ground truth rather than floating as raw URLs. Queries are classified with both keyword and embedding signals, and ranking runs on domain pagerank.
The detail that makes it lab-native: every source gets an agent-friendliness verdict, a judgment on whether a page will actually cooperate with a machine reader. That's what makes it the "phonebook of the AI-first internet": not just what's out there, but what an agent can actually use.
Highlights
- Built for agents, not eyeballs: a typed API instead of a results page
- 32M+ pages ingested from a Common Crawl-derived corpus
- Entity-graph grounding: the corpus is cross-referenced into the 4ort.xyz knowledge graph
- Smart query handling: keyword + embedding query classification
- Domain-pagerank ranking for result ordering
- Agent-friendliness verdicts per source, making it the phonebook of the AI-first internet
Status
● Live core (/search shipped); corpus expansion staged.