Stay free if you only need 500 one-time credits and access to /scrape, /crawl, /map, /search, /extract. Upgrade if you need 500,000 credits per month and 100 concurrent browsers. Most solo builders can start free.
Why it matters: Per-credit pricing escalates quickly for full-site crawls of large domains — a 100k-page crawl can exhaust a Hobby plan in a single run
Available from: Hobby
Why it matters: Free tier is capped at 500 credits with strict rate limits, making it useful for evaluation but not sustained development
Available from: Hobby
Why it matters: Highly dynamic, captcha-protected, or unconventionally structured sites can still produce imperfect markdown that requires post-processing
Available from: Hobby
Why it matters: Self-hosted version omits the managed proxy network and top-tier anti-bot measures, so cloud and self-hosted are not feature-equivalent
Available from: Hobby
Why it matters: Structured extraction quality depends heavily on schema/prompt design — naive schemas on complex pages yield inconsistent JSON
Available from: Hobby
Firecrawl provides reliable web-to-markdown conversion with JavaScript rendering and intelligent content extraction, with results typically returned in under one second. The crawl endpoint handles large site indexing via asynchronous batch jobs with webhook callbacks, automatic retries on transient failures, and configurable concurrency limits. The Standard plan and above include priority support SLAs, and the open-source self-hosted option lets teams run Firecrawl within their own infrastructure for maximum uptime control.
Yes, Firecrawl is open source under Apache 2.0 with 30,000+ GitHub stars and a documented Docker-based self-hosted deployment. The self-hosted version includes the core /scrape, /crawl, /map, /extract, and /parse endpoints with full functionality. The main trade-off is that self-hosted deployments do not include the managed proxy network and premium anti-bot measures available on the cloud service, so sites with aggressive bot detection may require additional proxy configuration when self-hosting.
Firecrawl charges per page scraped, with paid plans starting at $19/month for the Hobby tier. Optimize by using the /map endpoint first to discover URLs cheaply before committing credits to /scrape or /crawl on the pages you actually need. Set crawl depth limits and URL filters to avoid indexing irrelevant pages. For very high-volume use cases exceeding 500,000 pages per month, consider the Enterprise plan for custom pricing or self-host the open-source version to eliminate per-credit costs entirely, paying only for your own infrastructure.
Migration risk is unusually low for an AI infrastructure product because Firecrawl is open source — you can always self-host the same engine you were paying for. The API surface is small (URL in, markdown or JSON out), so switching to or from Firecrawl involves minimal code changes. Data portability is inherent since Firecrawl processes public web content on demand rather than storing proprietary datasets, and the Apache 2.0 license ensures no vendor lock-in on the codebase itself.
A custom Playwright stack gives you maximum flexibility but you become responsible for browser pools, residential and datacenter proxy rotation, anti-bot evasion, captcha handling, content extraction logic, and ongoing maintenance as websites change their structures. Firecrawl abstracts all of this behind a single API call that returns clean markdown. For teams whose core product is AI rather than scraping infrastructure, Firecrawl typically saves weeks of engineering time and delivers more reliable results across the long tail of website structures compared to maintaining a custom solution.
Start with the free plan — upgrade when you need more.
Get Started Free →Still not sure? Read our full verdict →
Last verified March 2026