How to get a Humaniser API key

Humaniser exposes a small API for products that need to check generated content before publication, rewrite it into cleaner editorial prose, or keep an audit trail of AI-content decisions.

What the API is for

The common workflow is simple: clean obvious machine phrasing, detect whether the text still reads as AI-generated, and only then run an editorial rewrite when the detector says the risk is high.

postprocess -> detect -> humanise(editorial)

The adversarial endpoint is not a bulk publishing tool. It is a lab endpoint for R&D and expensive manual tests.

Request a token

Ask the Humaniser operator for a client token and include your product name, environment, expected request volume, and whether you need detect, humanise, or both. The operator will issue a token scoped to a stable client id such as chargeintel or klarads.

Store the token as a server-side secret named HUMANISER_API_KEY. Never ship it in browser JavaScript, mobile apps, screenshots, logs, or URLs.

ChargeIntel setup

ChargeIntel should store its token server-side and call the production API at https://api.humaniser.eu. The token goes in X-API-Key, not in a query string.

HUMANISER_BASE_URL=https://api.humaniser.eu
HUMANISER_API_KEY=hum_chargeintel_REPLACE_WITH_ISSUED_TOKEN

Detect example

const res = await fetch("https://api.humaniser.eu/v1/detect", {
  method: "POST",
  headers: {
    "content-type": "application/json",
    "X-API-Key": process.env.HUMANISER_API_KEY,
  },
  body: JSON.stringify({
    text,
    threshold: 0.5,
    min_words: 50,
  }),
});

const verdict = await res.json();
if (verdict.confidence_band === "high_confidence_ai") {
  // block auto-publish or send for editorial review
}

Humanise example

const res = await fetch("https://api.humaniser.eu/v1/humanise", {
  method: "POST",
  headers: {
    "content-type": "application/json",
    "X-API-Key": process.env.HUMANISER_API_KEY,
  },
  body: JSON.stringify({
    text,
    mode: "editorial",
    channel: "eu_public",
    post_process: true,
  }),
});

const { rewrite } = await res.json();

Recommended integration tiers

/v1/postprocess first: deterministic cleanup, no model cost.
/v1/detect second: decide whether content needs review.
/v1/humanise only when useful: editorial rewrite.