Krenzo
Use cases

Wherever a model needs to know what's true right now.

The common thread: a system that reasons well but can't see past its training cutoff. Krenzo is the eyes.

Research agents

Multi-step agents that need to read before they reason. Fan a plan out into a dozen parallel searches, pull the sources, and hand the model a clean brief instead of a pile of HTML.

`depth: deep` + `time_range` keeps long research runs current.

RAG over the open web

Your vector store covers internal documents. Krenzo covers everything else — so a question about a competitor, a regulation, or this morning's news doesn't fall through the gap.

Returns pre-chunked passages sized for embedding.

Customer-facing chat

Support and sales assistants that answer from live documentation and pricing pages rather than a six-month-old snapshot. Citations render as links your users can check.

`quick` depth returns fast enough for a typing indicator.

Market and competitor monitoring

Scheduled jobs that watch a defined set of domains for changes, extract the substance, and push a summary into Slack or a dashboard.

`include_domains` scopes every run to sources you trust.

Due diligence and compliance

Pull filings, press releases, and regulatory notices for a named entity, with a provenance trail for every claim. Refusal-on-thin-evidence matters more than coverage here.

Answer API declines rather than speculating.

Content and SEO tooling

Analyze what already ranks for a topic, extract structure and claims, and ground a draft in real sources instead of model memory.

Batch extract handles a whole SERP in one call.

What teams tell us matters

Feedback from early design partners, paraphrased into the three things that come up every time.

Token cost, not API cost

The search bill is rarely the problem — the model bill for stuffing raw pages into context is. Trimming results is the whole game.

Latency budget is fixed

An interactive agent has roughly a second before it feels broken. Retrieval has to fit inside that, including the model's own thinking time.

Citations are non-negotiable

Anything customer-facing needs a link back to the source. An answer without provenance can't be shipped in a regulated context.

Which one is yours?