Firecrawl
by Firecrawl
Turns websites into LLM-ready markdown or structured JSON: scrape one page, crawl a site, map its URLs, search the web, or let an agent gather data.
Endpoints
- Core: Search, Scrape and Interact (browser actions such as click, scroll and type)
- Others: Parse (PDF, DOCX), Agent, Crawl, Map, Batch Scrape, Monitor (page change checks)
- Outputs: markdown, structured JSON, screenshots
- MCP:
firecrawl-mcpruns through npx with an API key; see its own repository
Licensing
The repository is primarily under AGPL-3.0; the SDKs and some UI components are under MIT, with the directory LICENSE files deciding. The hosted cloud version includes features beyond the self-hosted one. The GitHub page shows about 200,000 stars.
Billing model
Credits: scrape and crawl cost one credit per page, JSON extraction adds credits, search is billed per ten results, and Interact is billed per browser minute. Failed scrapes with no result are not charged, though error-status pages such as 403 and 404 still cost a credit.
| Tier | What it unlocks |
|---|---|
| Free | Monthly credits with no card, 2 concurrent browsers, community support, no pay-as-you-go |
| Hobby | 5 concurrent browsers, pay-as-you-go top-ups |
| Standard | 25 concurrent browsers, Data Processing Agreement |
| Growth | 50 concurrent browsers, priority support, Slack support |
| Scale | 100 concurrent browsers, one month of credit rollover, invoiced billing, custom analytics dashboard |
| Enterprise | SLA, custom limits, SSO, SCIM, Zero Data Retention, API-key IP restrictions, static egress IPs |
Amounts are on the pricing page.
Position
Firecrawl is the extraction end of the stack: where exa-search-api, parallel-search-api and brave-search-api find pages, Firecrawl fetches and cleans them, and it also offers its own search. It overlaps with browser-use and Bright Data (brightdata) for harder, interactive pages.