--- title: "Scraptio Integration" description: "Scraptio extracts the readable text of a web page, optionally narrowed to chosen elements. Workflows pull article bodies without navigation and footer noise, validate the key, and check usage." url: https://flowrunner.ai/integrations/scraptio date_modified: 2026-09-04T14:59:52-07:00 --- # Scraptio [Developer Tools](https://flowrunner.ai/integrations/category/developer-infrastructure) Scraptio extracts the readable text of a web page, optionally narrowed to chosen elements. Workflows pull article bodies without navigation and footer noise, validate the key, and check usage. [Verified](https://flowrunner.ai/integrations/verified "What does verified mean?") · 3 actions · API key · available [Scraptio website](https://scraptio.com/) · [Platform Documentation](https://scraptio.com/) · Capability data verified 2026-08-25 1. A schedule hands the agent this week's list of article URLs to ingest 2. Get Usage reports how many pages remain in the period and when the quota renews 3. Scrape URL pulls each page's readable text, narrowed to the element that holds the article body 4. Agent measures each result against the length the filter usually produces 5. Returns that came back near-empty or at full-page size are set aside as suspected filter failures 6. Clean article text lands in the knowledge base the rest of the workflows read from 7. A content owner opens the set-aside URLs and decides whether the filter broke or the article is gone ## What This Integration Enables Scraptio is three operations. There is no proxy pool, no browser engine, no job queue, and no crawler. It answers one question: what does this page actually say. That narrowness is the product, not a gap in the connector, and it is why Scraptio belongs in a workflow where something downstream is going to read the result closely. The expensive part of a retrieval pipeline is rarely the fetch. It is the boilerplate nobody stripped. A page of raw HTML carried into a model prompt spends most of its tokens on a navigation menu, and the answers that come back cite a cookie banner. FlowRunner agents call Scrape URL with a CSS filter, get the article body, and hand a clean block of text to the next step. The [human-in-the-loop](https://flowrunner.ai/concepts/human-in-the-loop) discipline earns its keep on the one thing Scraptio genuinely cannot tell you: whether an empty result means the page was empty or the filter stopped matching. ### Without FlowRunner **Boilerplate in the corpus**: Navigation, cookie notices, and footer links get indexed with the article and surface later as answers **A parser per publication**: Every new source means another selector script somebody owns and nobody remembers **Ingestion that fails quietly**: A page that returned nothing looks the same in the log as a page that ingested correctly ### With FlowRunner **Article text only**: A prefixed CSS filter narrows each fetch to the element that actually holds the writing **Configuration instead of code**: The filter is a field on a workflow step, versioned with the workflow rather than in a separate repository **Empty results reach a person**: Returns that land near zero or near full page are routed for review instead of into the index ## 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](https://flowrunner.ai/integrations/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](https://flowrunner.ai/integrations/notion) database with its source and fetch date. A digest posts to [Slack](https://flowrunner.ai/integrations/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](https://flowrunner.ai/integrations/anthropic-ai) or [OpenAI](https://flowrunner.ai/integrations/openai-ai) 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. ## Human-in-Loop Highlight Scraptio reports a filtered scrape that matched nothing as a success, and it has two failure shapes that look like opposites. A filter that no longer matches returns almost no text. An unprefixed filter matches nothing at all, and the vendor falls back to returning the entire page. So the agent checks the size of what came back rather than the status code: a result near zero and a result near the full page both mean the filter stopped working. When either shows up, the ingestion pauses on those URLs and the agent posts to the content owner: "Three sources returned outside their usual range this run. `theverge.com/...` came back at 40 characters, `stratechery.com/...` came back at 11,000 with the site navigation in it. Filters unchanged since March. Redesign, or paywall?" The agent could have indexed all three and reported a clean run. What made the corpus trustworthy is that the two ambiguous results went to somebody who could open the page and look. Agent processes routinely Detects exception requiring judgment Clear match Continues automatically Ambiguous Routes to human via preferred channel Human decides Agent resumes with decision ## Agent Capabilities 3 actions ### Extraction 1 - **Scrape URL** Returns a page's readable text, optionally narrowed to chosen elements. This is the whole product, and the Filters field is what makes the output usable: without one you get the page including its navigation and footer, and with a CSS class you get the article bodies. Every call spends a credit. ### Account 2 - **Validate API Key** Checks that the configured key works. Spends no scraping credit, which makes it the right connection test to run at the top of a flow rather than discovering a bad key partway through a batch. - **Get Usage** Returns the plan, the requests remaining in the current period, and the renewal date. The guard to put in front of any bulk run, because exhausting the quota mid-loop stops the execution rather than returning what it had. ## Frequently Asked Questions ### What can FlowRunner do with Scraptio? FlowRunner agents can run Scrape URL, Validate API Key, and Get Usage in Scraptio. ### Does connecting Scraptio to FlowRunner require OAuth? No. Scraptio connects to FlowRunner with an API key, no OAuth flow required. ### Can Scraptio trigger a FlowRunner workflow automatically? Scraptio doesn't currently expose triggers in FlowRunner. It connects as an action step inside workflows started by another trigger. **Work at Scraptio?** This integration exposes Scraptio to AI agents on every FlowRunner plan, including through MCP, at no cost to you. [See what FlowRunner offers integration partners](https://flowrunner.ai/integrations/partners), including how to keep this page current. --- Markdown version of https://flowrunner.ai/integrations/scraptio. Site index: https://flowrunner.ai/llms.txt