Turn the web into useful intelligence.

Create an account and choose how you want to use CrawlKit. You can start by describing what you sell to find potential customers, or create an API key and start working with web data directly. You can use both workflows from the same account.

Tell CrawlKit what you sell or provide your website. CrawlKit searches the web for relevant companies, researches their public websites and other relevant sources, then surfaces prospects that appear to fit your product along with the reasoning and evidence behind each result.

No. CrawlKit is designed to research the live web rather than simply filter a static company list. It can use information from company websites, product pages, documentation, careers pages and other public sources to understand why a company may be relevant to what you sell.

Yes. Customer discovery is optional. You can continue using CrawlKit purely as web infrastructure through Search, Scrape, Extract, Crawl, Watch and MCP.

Use Search when you know what information you need but do not already have the URLs. Search helps you discover relevant pages and sources before using Scrape, Extract or other CrawlKit workflows.

Scrape returns page content as Markdown, HTML, text, metadata, links or a screenshot. Extract returns structured JSON from a schema or instruction. Schema-based output is validated before a request succeeds.

Use Crawl to discover and fetch multiple pages from a website. Set page and depth limits, include or exclude paths, and retrieve the results as the job progresses. The default is 100 pages at depth 2, with a maximum of 1,000 pages.

Watch checks a URL on a schedule and compares normalized content or extracted fields. The first successful check creates a baseline. Later changes produce before/after data and a diff, with optional signed webhooks. Checks default to hourly; the minimum interval is 15 minutes. Change history is kept for 30 days.

Each successful page costs 1 credit, plus 3 if AI is used. Search costs 1 credit per 10 returned results, rounded up; empty results cost 0 credits. Crawl is charged per successful page. Each successful Watch check uses the same rate, even when nothing changes. Screenshots are included. Unused reservations are refunded.

Credits never expire. Start with 100 free credits and buy more when you need them. There is no monthly subscription requirement, and larger credit packs reduce the price per credit.

No. Standard scraping does not require an LLM. Extraction can read suitable embedded data directly; an LLM can handle instruction-based or other schema requests when configured. You can test your target page in the Playground.

Yes. The MCP server exposes Search, Scrape, Extract, Crawl and Watch tools to compatible clients. You can also call the HTTP API directly with your API key in the Authorization header.

HTML pages, JavaScript rendering and bounded PDF parsing are supported. Scanned PDFs can use optional English or Turkish OCR. Some protected sites may still reject requests, and scrolling remains bounded by configured steps and time. Test the pages you need before building a workflow around them.

Crawl returns a job ID. Scrape and Extract can also return HTTP 202 while work continues. Poll the job endpoint until it completes or fails.