Tisane Labs is a natural-language API for moderating user-generated content, and its output is what distinguishes it. Rather than a toxicity score, a request returns the exact offending span, its offset and length, a type such as personal_attack, a severity, and a plain-language explanation of why it was flagged.
That explainability is a deliberate position – the vendor is pointed about it, arguing Shapley values do not constitute explainability – and it is framed against DSA compliance, where a platform may have to justify a moderation decision. The second focus is adversarial text: asterisked words, algospeak, control characters and character substitution defeat rule and blocklist approaches, and handling them is the product’s core claim. Beyond abuse detection it covers entity extraction, granular topic modelling, language identification, built-in translation and multilingual word lists, with per-context switches for cases like adult advertising in profiles, in-game text or username parsing.
It ships ready-made integration plugins and a documentation portal, including a Perspective API migration guide. Signup is self-serve; specific pricing is not published on a public page.
