Firecrawl SDK

Use this skill when Firecrawl must live inside product code: the application itself calls the Firecrawl SDK or REST API at runtime, driven by a FIRECRAWL_API_KEY from its environment. This is the app-integration path, not the agent's own live web work — for ad-hoc scraping, crawling, searching or interacting done by the agent during a task, use the firecrawl CLI skill instead.

Typical product cases: knowledge ingestion from known URLs, enrichment from company or docs pages, answer generation with cited sources, extraction behind login-gated or JavaScript-heavy pages. The guides cover the three endpoints in build order: /scrape for one known page, /search when only a query is known, /interact when the page must be acted on before it can be read. Start with the cheapest endpoint that fits and escalate only when the data is not reachable yet: scrape → search for discovery → interact for pages that must be manipulated.

Guides

Open these in order of need, not all at once:

• overview.md — start here: scope of the SDK integration path and how it differs from the CLI skills. Read it first in every new integration.
• onboarding.md — install, browser auth, and wiring FIRECRAWL_API_KEY into the project's .env. Open it when the app has no Firecrawl access yet, or when the human still needs to sign up or authorize in the browser.
• scrape.md — content from one known URL: formats (default markdown), onlyMainContent for article-like pages, rendering options, and when to escalate to /interact. The default extraction primitive.
• search.md — the feature starts from a query, not a URL: turning a search into a shortlist of pages for later extraction. Keep search and extraction conceptually separate unless scraping results is clearly required.
• interact.md — pages that need clicks, typing, forms, pagination or navigation before content is readable. Keep it scoped to the smallest browser workflow that unlocks the data.

Endpoint order at a glance

• Have a URL, need page content → /scrape (scrape.md).
• Have only a query, need to discover sources → /search (search.md), then scrape the shortlist.
• Content locked behind clicks, forms or login → /interact (interact.md), scoped to the smallest workflow that unlocks the data.
• First Firecrawl contact in this app at all → onboarding.md before any of the above.

References

• references/project-setup.md — wiring Firecrawl into the project structure and environment.
• references/auth-flow.md — the browser sign-in / authorization flow when the human must approve access.
• references/sdk-installation.md — installing the chosen SDK or preparing the REST path.