API Documentation

Scan is open now. Integrate, Publish and Monitor will be documented here when they open.

Scan a page

One public page per request, no key needed. nectr reads the site's robots.txt, then fetches the page twice: as plain HTML, and in a headless browser that runs its JavaScript. It also checks up to 3 PDFs the page links to on the same site. nectr keeps neither the page nor the result: it goes back to you only.

curl -X POST https://nectr.ch/api/v1/scans \
  -H "Content-Type: application/json" \
  -d '{"url": "https://yoursite.ch/pricing"}'

The body has one field, url: any public http(s) page, up to 2048 characters. A scan takes a few seconds.

Response

Field	Type	Meaning
url	string	The page that was scanned.
title	string \| null	The page's title.
rendered_words	integer	Words in the page as markdown, with JavaScript.
plain_words	integer	Words in the page as markdown, without JavaScript.
rendered_markdown	string	The page with JavaScript, as markdown.
plain_markdown	string	The page without JavaScript (its HTML), as markdown.
only_with_javascript	string[]	Blocks of the rendered page whose words are nowhere in the plain HTML. At most 50, each cut at 300 characters.
missing_word_share	number	Words in those blocks over words in the rendered page, from 0 to 1.
robots_found	boolean	false when the site has no robots.txt. Every crawler is then allowed.
robots	object[]	`{agent, allowed}` for GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, PerplexityBot, Google-Extended and CCBot: whether the site's robots.txt lets each one fetch this page.
pdfs	object[]	Up to 3 PDFs the page links to on its own site, in page order. Each one is described in the next table.
pdfs_not_checked	integer	Linked PDFs left out: past the first 3, or on another site.
scanned_at	string	When the scan ran, in ISO 8601.

Each entry in pdfs:

Field	Type	Meaning
url	string	The PDF's address.
pages	integer \| null	Its page count.
words	integer \| null	Words in its first 5 pages at most.
has_text	boolean \| null	false when it has no text layer (fewer than 5 words a page): probably a scanned image. null when it wasn't checked.
note	string \| null	Why it wasn't checked, when it wasn't.

Example

example.com, shortened. Its HTML holds one English paragraph, and a script adds five translations, so 85% of the text only appears with JavaScript.

{
  "url": "https://example.com/",
  "title": "Example Domain",
  "rendered_words": 133,
  "plain_words": 25,
  "rendered_markdown": "This domain is for use in documentation examples …",
  "plain_markdown": "This domain is for use in documentation examples …",
  "only_with_javascript": [
    "L’usage de ce domaine est réservé à des exemples de documentation …",
    "…"
  ],
  "missing_word_share": 0.85,
  "robots_found": false,
  "robots": [
    { "agent": "GPTBot", "allowed": true },
    { "agent": "ClaudeBot", "allowed": true },
    "…"
  ],
  "pdfs": [],
  "pdfs_not_checked": 0,
  "scanned_at": "2026-10-09T20:31:12Z"
}

Errors

Errors are JSON with a detail message you can show as it is. Two exceptions: a malformed body gets a 422 whose detail is a list of problems, and a 429 has an error message instead.

Status	When
403	The site's robots.txt asks Nectrbot not to fetch this page.
422	The URL isn't a public http(s) address, it's a document such as a PDF, the page answered with an HTTP error or couldn't be loaded, or the site's robots.txt couldn't be read.
429	More than 10 scans in a minute from one address.
503	nectr is busy. Try again in a minute.

robots.txt

nectr's crawler is Nectrbot. A page your robots.txt disallows for it isn't fetched, and neither are PDFs it disallows.
A missing robots.txt (a 4xx answer) allows everything.
A robots.txt that can't be read (a 5xx answer, or none) stops the scan, as RFC 9309 asks.

Accounts and API keys

Scans need no account or key.
New accounts are paused, so new API keys can't be created.
Existing API keys keep working for now. The latest update gives the dates.

The MCP server

nectr-mcp keeps working with existing API keys. When the new tools open, it will move to them: checking how AI reads your own site, and serving it.

Questions? Get in touch.