Browse.ai's whole model is train-by-pointing: install the Chrome extension, open the site, and tell the robot what to capture - a list, a text block, or a screenshot. For lists you highlight the list, double-click each field, and label them (title, price, colors). Separate robot type for change monitoring on an hourly-to-weekly schedule. If you are trying to script Browse.ai like an API, you are fighting its design - the training happens in the browser, the API is for running and reading results.

Context: Browse.ai guides: there are two robot types - Extract Structured Data and Monitor Site Changes. For extraction, you train the robot in the browser with the Chrome extension: choose to capture a list, text, or screenshot. For lists, highlight the list on the page, double-click the text you want, and assign each field a category (e.g. product title, price, colors). Browse.ai then generates the structured list in seconds. Monitoring robots check on a schedule from every hour to once a week and notify on changes.