Firecrawlv1.0.0

An AI agent researching a topic tries its free web_fetch tool first, and when pages come back empty because they're JS-heavy SPAs, it escalates to `fc search "<topic>" --limit 5 --scrape` and parses the JSON to assemble a source-backed brief with full page content rather than snippets.

A support agent maps a vendor docs site with `fc map https://docs.example.com --limit 100`, selects relevant URLs, then scrapes or bulk-crawls the tree into a local mirror for retrieval. This gives the bot accurate, up-to-date documentation content instead of stale embeddings.

An agent periodically scrapes competitor product pages that render prices client-side, extracting clean markdown to track price changes, stock status, and promotional copy. Using `fc scrape --out` keeps a per-run archive for diffing against previous snapshots.

A compliance agent uses the direct REST `scrape` endpoint with `actions` (wait, click, write) to log in to a portal, navigate to a report page, and return its markdown. This handles interactive, authenticated flows the CLI doesn't cover yet.

A strategy team runs a limited `fc crawl https://blog.example.com --limit 50 --out ~/mirror/` to pull a competitor's full blog archive for theme analysis, being mindful of credit spend before large `--limit` runs.

Firecrawl charges per operation, where each scrape or search result costs about 1 credit and each crawled page costs 1 credit, with a 500-credit/month free tier. This pay-as-you-go model aligns cost with the volume of web data an agent consumes.

Instead of selling to end users, Firecrawl positions itself as backend infrastructure that other AI products embed to render JS-heavy pages, perform full-content search, and map or crawl sites without maintaining a headless browser fleet. Value is captured from developers building agent products on top of it.

Free tier of 500 credits/month and a low-friction CLI and REST API lower onboarding cost so individual developers adopt it easily, then hit limits and convert to paid plans. The skill even instructs agents to warn users before large crawl or map calls, nudging upgrade decisions.