Web scraper (Firecrawl)

Runs on its own 12

Looking things up, writing to Everdeck's own records and changes one step can undo. Once an agent's charter grants one of these, it runs without asking, unless you set it to Ask first in the charter or one of your approval rules covers it.

• Read a web page Tier 0 · Read Outside content

  Fetch a page or PDF and return its main content as Markdown, with its title, final URL and HTTP status. Add summary to formats for a short summary, or links for the page's links. Handles JavaScript-heavy pages; set wait_ms when content appears late. A cached copy up to 2 days old is used unless max_age_ms is lower (0 forces a fresh fetch).

• Ask a question about a page Tier 0 · Read Outside content

  Read a page and answer one question about it, for example "What is the refund window?" or "Which plans include SSO?". Returns only the answer, so it suits long pages.

• Extract structured data from a page Tier 0 · Read Outside content

  Pull specific facts from a page as JSON, for example a product's name, price, currency and stock status, or a company's contact details. Describe what to extract; add a JSON Schema to fix the result's shape.

• Search the web Tier 0 · Read Outside content

  Search the web (and optionally news or images) and return titles, URLs and descriptions. Use search_and_read when you also need the pages' text.

• Search and read the top results Tier 0 · Read Outside content

  Search the web and return the top results with each page's main content as Markdown. Costs a page read per result on top of the search, so keep limit small.

• List a website's pages Tier 0 · Read Outside content

  List the URLs on a website (from its sitemap and links) with titles where known, without reading them. Give search to rank URLs by relevance, for example "pricing" or "blog".

• Crawl a website Tier 2 · Reversible

  Start reading many pages of a site in the background (for example a help centre or a competitor's docs). Returns a crawl id; check it with get_crawl. Each page costs a credit, so the limit is capped at 100 pages per crawl. Use include_paths such as ["blog/.*"] to stay focused.

• Check a crawl and read its pages Tier 0 · Read Outside content

  Get a crawl's status (scraping, completed, failed or cancelled), page counts, credits used and the pages read so far as Markdown. Large results arrive in parts of about 10 MB; when next is present, ask for the next page to get the rest.

• Watch pages for changes Tier 2 · Reversible Outside content

  Check up to 20 pages on a schedule and start agent work when one changes, appears or disappears. Give a goal (for example "price or plan changes") so only meaningful changes count. Every check costs a credit per page; the result shows the estimated credits per month.

• List watched pages Tier 0 · Read Outside content

  List the page monitors on the Firecrawl account with their schedule, status, pages and estimated monthly credits.

• See what changed in a check Tier 0 · Read Outside content

  Read one monitor check: the pages that changed, appeared or disappeared, each with a unified diff and, when a goal is set, whether the change is meaningful and why.

• Pause or resume watching Tier 2 · Reversible

  Pause a page monitor (no more checks or credits) or resume it.