Classify any text into your own categories — intent, topic, routing — with no training data.
This text classifier sorts a piece of text into categories you define — no training data, no labeled examples, no model to maintain. Type the text, give it two to twenty labels, and it returns a score for every label plus the best match and a one-sentence rationale naming the evidence it used. Because the labels are yours, the same tool works for intent detection (cancel, upgrade, question), support-ticket routing (billing, shipping, technical issue), topic tagging, lead qualification, content moderation buckets, or custom sentiment scales.
By default exactly one label applies and the scores form a distribution that sums to 100%. Tick Multiple labels can apply when a text can legitimately belong to several categories at once (the example above is both a billing problem and a technical issue): each label then gets an independent score, and every label at or above the threshold is selected. In that mode it is valid for no label to be selected — "none of these apply" is often the right answer, and a catch-all Other label is a good habit either way.
The optional Context field tells the classifier what the texts are or how to decide borderline cases (for example "Support tickets for a SaaS billing product; prefer Billing when a charge is mentioned"). It steers the decision but can never add a label outside your list. Text is treated as data, so instructions inside it are ignored.
Need this in your own product or pipeline? The Classify API takes the same inputs, plus an optional short description per label to sharpen the decision, and returns the same scores, selected labels and rationale as JSON. Pair it with the sentiment and tone utilities, or contact us about routing and tagging at volume.