Skip to main content
targets says what a monitor watches. It holds 1 to 50 targets, and each has a type of scrape, crawl or search. One monitor can mix types, and every check reads every target.

Make a request

Get an API key in the Quickstart. This request watches one page and judges each change against a goal. The page returns a new identifier on every read, so a check always finds a change.

Read the response

This real response is cut to the fields this page is about. Stophy added an id to the target and turned the judge on because you sent a goal.

Scrape targets

A scrape target lists the pages to read on every check. An address sent twice is read once.

Scrape options

scrapeOptions takes the same fields as Scrape, except url. A monitor always reads the page live, so it ignores maxAge and minAge. zeroDataRetention is refused, because a monitor keeps a snapshot of each page to compare with the next. formats says what to compare. The first of markdown, html, rawHtml and links is compared, and markdown is the default. A page counts as changed when that text differs from the last read. To compare fields instead of text, add a changeTracking format or a json format with a schema or a prompt, as in Change tracking and JSON. The modes of changeTracking are git-diff, which returns the text diff, and json, which returns each field before and after. With only json, a page whose text changed but whose fields did not is same. json adds $3 per 1,000 pages when it returns fields.

Crawl targets

A crawl target starts from one page. Every check finds the pages under it again and reads each one, so a page that appears later shows up as a new change.
This real target, set on a monitor with PATCH /v1/monitor/{id}, was stored with every default filled in:
The check that ran afterwards read one page, because the site has no other page under that address, and cost $0.001. A crawl target takes scrapeOptions too. If a check finds no pages, it fails and says so in error.

Search targets

A search target runs your queries on every check and reports results it has not seen before. It reads no page, so it has no baseline: a meaningful result is reported the first time it is found.
This real target was stored with excludeDomains set to [] and an id. A check with it returned three results and cost $0.004: $0.001 for the search and $0.001 for each result the judge read. A search target needs a goal, or judgeEnabled: false. Without either, the call returns 400 invalidRequest with goal: a search target needs a goal to judge results against, or judgeEnabled set to false. A result the judge finds meaningful is reported as new. A result it finds not meaningful is marked ignored and is not looked at again. A result you have already been told about is already_seen. See Checks and changes.

Goal and judge

goal is what matters to you, in plain words, up to 2,000 characters. With a goal, a model judges every changed page against it. judgeEnabled: false turns the judge off while keeping the goal. The judgment is on the page of the check and in the webhook payload.
confidence is high, medium or low. A change the judge calls not meaningful is still recorded and still sent, with meaningful: false, so you decide what to do with it. If the judge cannot decide, judgment is null and the page is not charged for it. Judging adds $1 per 1,000 pages for each changed page it reads.

Update targets

PATCH /v1/monitor/{id} replaces targets as a whole, so send every target you want to keep. You can also change goal and judgeEnabled in the same call.
  • A page you keep stays compared with its last snapshot.
  • A page you add to a scrape target is read for the first time on the next check, which only records it as new and sends no event.
  • A page you remove is no longer read or billed.

Limits and errors

  • A monitor takes at most 50 targets and 100 pages in one scrape target. Your account takes 25 monitors and 1,000 pages in all. See Limits and errors.
  • An address that is not a public website returns 400 invalidRequest with targets.0.urls.0: Expected a public website URL.
  • zeroDataRetention inside scrapeOptions returns 400 invalidRequest with zeroDataRetention is not available for monitors, which keep a snapshot to compare.
  • Send targets and schedule together. The older fields url, urls, interval, formats and changeTracking still work on their own: url or urls for the pages, interval for seconds or text such as 30m, 6h or 1d, and changeTracking for the compare modes. Sending them together with targets or schedule returns 400 invalidRequest.