One API. Honest defaults.
Most classification APIs fall into one of two buckets: zero-shot LLM wrappers that hallucinate labels, or rigid SaaS that locks you into a single taxonomy and a sales call. Classivore sits between them — stable taxonomies, swappable models, published rate cards.
Who builds this
small team · no bot support
Classivore is operated by Noe Software Solutions LLC, a Connecticut limited liability company. Our flagship models are trained in-house, and every classification runs on infrastructure we operate. Nothing passes through to a third-party API.
Built by Mike Noe — 20+ years in ad tech, most recently CTO of a contextual/RTB ad-tech company acquired by Azerion in 2025. Independent since 2026.
The team is small on purpose. One person answers support@classivore.com, reviews every pull request, and responds to bug reports. When you email, a human reads it.
Why it exists
built because nothing fit
We needed a classification API that did four things at once: returned the same response shape across every model, didn't require prompt engineering, published its rate card instead of hiding behind “contact sales,” and charged for what you actually use rather than bundling ten features you won't touch.
Nothing on the market checked all four. So we built it.
The first production model is the IAB Content Taxonomy 2.2 — the ad-tech industry standard with 698 categories. It's a fine-tuned DeBERTa-v3-large classifier trained on web content we collected and labeled ourselves, not a zero-shot LLM wrapper. More taxonomies are coming.
The training pipeline is open source: github.com/NotYoCheese/classivore.
Notes from building this
how things actually work, written down
Get in touch
one inbox · one business day