searchsearch + scrapemonitor create / list / getPurchased top-up credits never expire and are not tied to your current billing cycle. Cancelling your subscription zeroes only the plan's included/monthly credit allowance; any purchased top-up balance survives cancellation and stays spendable.
The FREE plan grants 1000 credits once, for the lifetime of the account, not monthly, and it never resets. No card is required to get them. Once they are spent, a FREE account either upgrades to a paid plan or stops.
+The FREE plan grants 500 credits once, for the lifetime of the account, not monthly, and it never resets, plus up to 500 more from onboarding tasks. No card is required to get them. Once they are spent, a FREE account either upgrades to a paid plan or stops.
The crawl start reserves the job. Subsequent polls charge only for newly materialized pages, not for the total accumulated page count each time.
That prevents the same already-seen pages from being charged again and again just because you are checking progress.
@@ -383,15 +383,15 @@Available tags: latest, 0.16 (tracks the current minor), 0.16.0 (pinned).
Use a pinned tag in production — latest rolls forward on every release.
git clone https://github.com/us/crw.git
cd crw
docker compose upTurn websites into usable data with one API. Start with a single scrape request, then move into search, map, crawl, extract, or MCP only when your workflow actually needs them. Interactive browser automation is handled by the companion crw-browse service.
Open-source web scraping API for agents and RAG. Scrape, crawl, search, map, and extract data from any site with one Rust binary, self-hosted or on fastCRW Cloud.
npx crw-mcp@latest install to add the MCP server and skill, then use the crw_scrape, crw_search, crw_map, and crw_crawl tools (or the native /v1 REST API) to fetch, search, and crawl the web with the CRW_API_KEY in my environment.
+:::
+
+Full reference written for autonomous agents: [Agent Onboarding Guide](/agent-onboarding).
+
+## Pick your endpoint
+
+:::cards
+::card{icon="code" title="Scrape" href="#scraping" description="One known URL in, clean markdown or JSON out."}
+::card{icon="search" title="Search" href="#search" description="No URL yet? Search the web and get ranked results with content."}
+::card{icon="map" title="Map" href="#map" description="Discover every URL under a domain before you scrape or crawl."}
+::card{icon="layers" title="Crawl" href="#crawling" description="Scrape every page under a URL in one background job."}
+::card{icon="key" title="Extract" href="#extract" description="Pull structured JSON fields from a page with an LLM."}
+::card{icon="plug" title="MCP" href="#mcp" description="Give Claude, Cursor, Codex, and other agents live web access."}
+:::
+
> **New to CRW? Use `/v1`.** The `/v1` routes are the native fastCRW API for new integrations. Use `/firecrawl/v2` when migrating existing Firecrawl v2 SDK code or when you need compatibility-only routes such as batch scrape or PDF parse.
https://api.fastcrw.comchunkStrategy is an optional request field on POST /v1/scrape that tells the engine to split the page's markdown into smaller pieces server-side before returning it. You supply an object with a type ("topic" to split on headings, "sentence" to split on punctuation, or "regex" for a custom pattern), an optional maxChars cap per chunk, and an optional dedupe flag to drop near-duplicate chunks. The engine returns the chunks as a chunks array alongside the normal markdown field — no text-splitting library needed on the client side. Note that chunkStrategy only works on POST /v1/scrape; it is not forwarded by batch or crawl jobs. See the RAG recipe for a complete example.
A credit is the billing unit for the hosted cloud at fastcrw.com, worth $0.001. Every scrape costs 1 credit regardless of renderer (HTTP, lightpanda, or Chrome); there is no renderer or proxy surcharge. LLM-backed work (POST /v1/extract, or formats: ["json"] / summary on scrape) costs that 1-credit base render plus the actual LLM cost of the call, dynamically metered: there is no flat credit number for it. See the Credit Costs table for current billing. Map and search start-calls each cost 1 credit, and crawl jobs charge 1 credit at start plus one additional credit per page as pages are discovered during polling. The FREE plan grants 1000 credits once, for the lifetime of the account, not monthly, and they never reset. Purchased top-up credits never expire and survive plan cancellation. Self-hosted deployments have no billing layer and are unaffected by credit costs. Check your balance with GET /api/v1/account/balance on the SaaS control-plane. See Credit Costs for the full table.
A credit is the billing unit for the hosted cloud at fastcrw.com, worth $0.001. Every scrape costs 1 credit regardless of renderer (HTTP, lightpanda, or Chrome); there is no renderer or proxy surcharge. LLM-backed work (POST /v1/extract, or formats: ["json"] / summary on scrape) costs that 1-credit base render plus the actual LLM cost of the call, dynamically metered: there is no flat credit number for it. See the Credit Costs table for current billing. A map call costs 1 credit, a search call costs 2 credits, and crawl jobs charge 1 credit at start plus one additional credit per page as pages are discovered during polling. The FREE plan grants 500 credits once, for the lifetime of the account, not monthly, and they never reset, plus up to 500 more from onboarding tasks. Purchased top-up credits never expire and survive plan cancellation. Self-hosted deployments have no billing layer and are unaffected by credit costs. Check your balance with GET /api/v1/account/balance on the SaaS control-plane. See Credit Costs for the full table.
"Firecrawl-compatible" means fastCRW exposes a /firecrawl/v2 compatibility layer for existing Firecrawl v2 SDK integrations. It is intentionally close enough for migration work, but it is not the recommended API for new fastCRW builds. New projects should start with /v1; migration projects should validate the documented differences before switching production traffic. See Migrate from Firecrawl for the exact list of changes and Compatibility for the behavior matrix.