Scrape Single Url

Agent calls POST /v1/scrape for a docs page and needs only clean markdown for an LLM, but passes formats=['markdown','html','rawHtml','screenshot'] 'to be safe.'

Request only the formats actually consumed — formats=['markdown']. Each extra format (especially screenshot) adds latency and consumes additional credits. Read response.markdown from the per-format payload; do not request rawHtml/screenshot unless a downstream step uses them.

A blog page returns markdown that includes the nav bar, cookie banner, and footer. The agent feeds the whole thing to an LLM and wastes tokens on chrome.

Set onlyMainContent=true so Firecrawl strips navigation, headers, footers, and boilerplate, returning the article body. Verify the returned markdown excludes nav/footer before treating token counts as content. Re-enable chrome only when the task specifically needs it.

Agent wants only the <article> body and must drop <aside class='ads'> elements from a content-heavy page.

Use includeTags (e.g. ['article']) and excludeTags (e.g. ['aside','.ads']) CSS selectors to scope extraction at scrape time. Confirm the selectors match the target site's DOM; pair with onlyMainContent where appropriate rather than post-hoc string deletion.