Use cases
Each page names the specific parameters that make the workload work, and a section on where this is not the right tool. No customer names — we do not cite customers without permission.
RAG over the open web
Retrieve, deduplicate, trust-rank and pack to a token budget — either as separate calls you control, or as one call that does the whole loop.
Teams grounding an LLM in current web content rather than a private corpus.
Autonomous research agents
Graph traversal replaces the search-read-search loop: triage by centrality, build entity briefs in one call, and follow edges rather than guessing queries.
Teams building agents that investigate a topic across many sources.
Fact-checking with corroboration counts
Count distinct hosts asserting a claim rather than counting copies — and get the count from an index where deduplication does not destroy it.
Teams that must verify claims before repeating them.
Monitoring topics and entities
Browse by recency with no query at all, bound by publication date, and collapse syndication so a monitoring feed carries distinct events.
Teams tracking coverage of a topic, company or event over time.
Competitive and market intelligence
Entity dossiers, co-mention discovery and connecting paths — the operations that make competitive research structural rather than a search loop.
Teams tracking companies, products and market movements.
Documentation search for coding agents
The code vertical and the docs content type remove tutorial blogspam; the term filter makes error-code lookup exact.
Teams building coding assistants and developer tooling.
Grounded question answering
Get a packed, corroborated context set in one call, with a reason attached to every passage so the answer can be explained.
Teams building assistants that answer factual questions from current sources.
Regulatory and compliance research
Time-bounded queries, authority filters, uniform corroboration thresholds and an endpoint that explains any exclusion.
Teams in finance, legal, pharma and the public sector who must defend a retrieval decision.
Finding sources worth following
Centrality surfaces what a field treats as foundational; similar and related widen from a seed without inventing new queries.
Teams building recommendation, curation and discovery products.
Serving users in many languages
One embedding space for 101 languages means retrieval crosses languages without a translation step or a per-language index.
Teams whose users search and read in different languages.
Grounding sovereign and public-sector agents
The evidence layer deployed inside your jurisdiction: no query leaves, the record of what was asked is yours, and the model reasoning over the evidence is your own.
Government, regulators, public broadcasters and defence programmes running agents on a model they control, inside a boundary they control.
What these pages are for
Each one takes a workload we know this API suits, states the problem the way the buyer would state it, and then names the specific parameters that make it work. They are not case studies. We do not have permission to cite customers, so there are no customer names and no invented metrics anywhere on them — what you get instead is the configuration and an honest section on where the workload stops being a good fit.
The parameter tables are the part worth reading twice. Most of the difference between a retrieval stack that works and one that quietly returns noise is four or five query parameters set deliberately rather than left at their defaults, and those defaults are tuned for a general case that is rarely yours.
The pattern underneath all of them
Three moves recur across every workload here. Filter before you fetch, because a result the agent reads and rejects costs as much as one it uses. Collapse near-duplicates at the source, because ten slots carrying one fact ten times is the most common way a context window is wasted. And treat absence as an answer rather than as silence — an agent that infers a thing does not exist from an empty result set is the single most reliable hallucination generator in a retrieval stack.
If your workload is not listed, the closest match is usually still the right starting point; the guides go deeper on the reasoning, and the glossarydefines anything here that is not obvious.
Start with the free tier
10,000 credits a month, no card. Every parameter documented here works on every plan.