Scrapfly review

Scrapfly is a hosted API platform, not a scraping library or self-hosted server. One API key provides access to endpoints for fetching public URLs, operating a managed Chromium browser, capturing screenshots, extracting structured data, crawling sites and running browser agents. The homepage also lists an MCP server for connecting compatible clients.

In a stack, Scrapfly sits between your application or agent and the websites it needs to read. Its managed services handle proxy rotation, browser execution and retries; your code sends requests to its APIs and consumes the returned data. The Web Scraping API can return HTML or markdown, while the Extraction API turns HTML into typed data using prompts, schemas or templates. Cloud Browser supports Playwright and Puppeteer compatibility over CDP. The Crawler API traverses sites and routes each page through the Web Scraping API. This reduces the infrastructure your team operates, but makes the scraping path dependent on a hosted vendor and its credit system.

The main integration surface is a set of hosted APIs. The homepage shows POST endpoints for scraping, screenshots, extraction, crawling and browser agents; Cloud Browser connects over the WebSocket CDP URL wss://browser.scrapfly.io/cdp?key=.... It describes a stable JSON envelope for API responses. The MCP server is available at mcp.scrapfly.io and is listed for Claude, Cursor, Cline, Windsurf and ChatGPT.

Scrapfly lists first-class SDKs for Python (pip install scrapfly-sdk), TypeScript (Node, Deno and Bun), Go (go get github.com/scrapfly/go-scrapfly) and Rust (cargo add scrapfly-sdk). It also offers a full-featured Scrapy extension. The homepage links documentation for each product and lists integrations including Zapier, Make, n8n, LlamaIndex, LangChain and CrewAI. No CLI is listed; confirm CLI support and the SDK methods you need in the vendor’s documentation. Check how retries, timeouts and response errors map to your application before shipping.

Scrapfly has a Free plan with 1,000 signup credits, access to every product, no credit card and no time limit. Discovery costs $30/month for 200,000 credits and five concurrent requests, but has no pay-as-you-go overflow. Pro is $100/month for 1,000,000 credits and 20 concurrent requests; listed overflow is $3.50 per 10,000 credits.

Startup costs $250/month for 2,500,000 credits and 50 concurrent requests, with overflow at $2.00 per 10,000 credits. Enterprise is $500/month for 5,500,000 credits and 100 concurrent requests, with overflow at $1.20 per 10,000 credits. Custom plans negotiate unlimited credits and committed concurrency; contracts start at $1,200/month and can scale above $30,000/month. The page lists monthly prices, says no annual contract is required and gives no annual rate. Scrapfly says credits scale with target complexity, so estimate consumption for each product and target. The free credits are a starting allowance, not a published recurring monthly quota.

Scrapfly is presented as a hosted service. Its homepage describes proprietary engines, Curlium and Scrapium, and a managed browser and proxy mesh; it does not list a self-hosted deployment option. The vendor does not publish an open-source licence or source availability for the service on the vendor's site, so teams that require auditable or modifiable code should ask Scrapfly directly.

The homepage lists SDKs and a Scrapy extension, but those client integrations do not make the hosted service portable. Your application can call its documented APIs, yet migrating means replacing scraping, browser and extraction behavior and checking how your data pipeline handles different response formats. Scrapfly does not publish data-export guarantees here. Confirm export formats, retention and deletion terms, and whether logs or extracted data can be retrieved before storing operationally important information in the service.

The pricing page sets explicit concurrency tiers: five requests on Discovery, 20 on Pro, 50 on Startup and 100 on Enterprise. Log retention ranges from one week on Discovery to four weeks on Enterprise. The homepage says failed requests are free and advertises a 99.99% success rate, but that is a vendor claim, not a service-level commitment. Confirm API rate limits, timeout behavior, regional availability and any SLA before relying on those figures.

Scrapfly says it manages retries and provides observability for success rate, throughput and latency. Its homepage describes a proxy mesh across 190+ countries and browser automation through managed Chromium. For production, check credit consumption against your mix of targets and products, and decide what happens when concurrency or plan credits are exhausted. Security, compliance certifications and detailed data-retention controls should be confirmed with the vendor. In one Reddit thread, a real-estate operator reported improved results after migrating about 80 targets and reduced costs versus traditional proxies; another commenter argued scraping tools can be a poor fit for unique targets. Those are individual reports, not independent benchmarks.