Choosing which checks run

Every category is a toggle on the request. Turn off what your product does not show; the pipeline skips that work, the report omits those findings, and the check costs less: each pass that is on adds its rate to the price (Credits and limits).

A key you leave out is on. Every option defaults to true, so { "plagiarism": true } on its own runs every check at the full 150 credits per 250 words, not a plagiarism-only check at 110. To run fewer checks, write out the ones you do not want as false, as the shapes below do. The cost calculator prints the exact options body for any combination. A key the API does not know (plagarism) is refused with 400 invalid_options naming it, so a typo never runs, or bills, every pass.

{
  "text": "…",
  "options": {
    "plagiarism": true,
    "grammarSpelling": true,
    "citations": true,
    "aiContent": true,
    "style": true,
    "register": "academic"
  },
  "citationStyle": "APA"
}

citationStyle (APA, MLA or Chicago, default APA) sets the style the citation findings and their fixes are written in. Each citation issue also carries messageByStyle and fix.byStyle with all three, so a product that lets the user switch styles can do so without a second check.

Common shapes

A category that is off is neither run nor charged, and a request with every category off is refused with no_checks_selected before anything is priced (while the style layer is switched off product-wide, style on its own counts as nothing on). Per started 250 words, a check costs 40 credits plus 70 for plagiarism, 20 for grammar and spelling, 10 for citations and 10 for AI content: the plagiarism checker shape above is 110 credits per 250 words, the AI detector 50, the writing assistant 60, and everything on 150. register and citationStyle change wording, not price. The 202 answer quotes the estimate as estimatedCredits; the finished check carries creditsCharged and the options it ran.

Prose, code and language

Documents can be prose, source code or both, in any of the supported languages. The pipeline detects each segment's language and searches accordingly; the report's languages and contentType (prose, code, mixed) say what it found. There is nothing to declare on the request.