Scraptio

Use Case Scenarios

• A reading list that stays current without a parser per publisher

  A team keeps a list of industry sources in Google Sheets, each row carrying a URL and the CSS class that holds that publication's article body. On a schedule the agent reads the sheet, calls Scrape URL once per row with the matching filter, and writes each result into a Notion database with its source and fetch date. A digest posts to Slack. Adding a publication means adding a row, not writing and testing another extraction script.

• Text a model can use, at a token bill somebody can defend

  Before a summarization step runs, the agent narrows each page to its article element. What reaches Anthropic or OpenAI is prose rather than markup, which cuts both the token count and the number of confident summaries built on a sidebar. The filter prefix matters here in a way that is easy to get wrong: an id needs a hash, a class needs a dot, and a tag needs angle brackets. An unprefixed value is not rejected, it simply matches nothing and the whole page comes back, which reads downstream as a very verbose article.

• The batch that stops before the credits do

  Running out of Scraptio credits mid-loop is an HTTP 402, and a 402 arriving on page 60 of 200 does not return the first 59 as a partial result. It stops the execution. So the agent calls Get Usage first, compares the remaining quota against the batch size, and either trims the run or schedules the remainder against the renewal date the same call reports. Validate API Key runs ahead of that as the connection check, because it spends nothing.