Classify website changes

When Wachete detects a change on a monitored page, have Claude judge whether the change is significant before anyone is alerted.

Try:

See all Wachete + Claude integrations or talk to a human

Trusted by teams already building on Knit·

4.9/5 on G2·

Google for Startups Accelerator

How Knit builds this workflow

The Agent researches both APIs and wires the trigger on one side to the action on the other — no pre-built connector required. Here is the vocabulary it has to work with.

Wachete Explore Wachete →

Triggers

Scrape completed

Fires when a scrape, crawl or robot run finishes in Wachete.

New result

Fires when Wachete returns a new batch of extracted data.

Website change detected

Fires when Wachete detects a change on a monitored page.

Actions

Scrape a URL

Extracts the content of a web page with Wachete.

Crawl a website

Starts a crawl of a whole site in Wachete.

Run a scraper

Runs a saved scraper, task or robot in Wachete.

Claude Explore Claude →

Triggers

New result returned

Fires when Claude returns a completion, prediction or other result.

Model run failed

Fires when a request to Claude errors or times out.

Fine-tune job completed

Fires when a fine-tuning job in Claude finishes.

Actions

Generate a completion

Sends a prompt to Claude and returns the model's response.

Run a model on an input

Sends text, an image or other data to Claude and returns the model's output.

Classify or extract data

Uses Claude to classify or extract structured data from text or images.

Summarize each scraped page When a scrape completes in Wachete, send the page content to Claude and store the summary it returns. See details

Extract structured fields from results When Wachete returns a new result, ask Claude to pull out the fields you need, such as product name and price. See details

See all Wachete + Claude integrations → · Browse the full workflow library →

Enterprise-grade security

Visit our Security Center Review the DPA

FAQ

Common questions

Can Knit build “Classify website changes” between Wachete and Claude?

Yes — describe it in the box above and Knit's Integrations Agent researches Wachete and Claude's public API docs (or your own uploaded docs) and builds a working workflow, whether or not either app already has a pre-built connector.

How long does it take to build?

Minutes to a first working version, not weeks — you test it against real data before it goes anywhere near production.

Why send scraped pages to an AI model?

Scraped HTML and text are noisy. Passing each Wachete result to Claude lets you pull out prices, names or key points in a consistent shape instead of cleaning the data by hand.

Can the model decide which pages to scrape next?

Yes. You can have Claude produce a list of URLs or search terms, then run a scrape in Wachete for each one.