Nine endpoints between a URL and clean data.
Crawl whole sites, scrape one page, search the web, capture screenshots, run AI extraction. Same Bearer auth, same JSON shape, optional streaming.
- Rate limit
- 10K rpm
- Core endpoints
- 9
- Output formats
- 4
- JSON & streaming
- REST
One API. Nine endpoints.
Each endpoint focuses on one shape of job. Switch between them with one parameter.
Discovery
Crawl
Recursively crawl entire websites and collect every page. Set depth limits, respect robots.txt, and get structured output in markdown, HTML, or plain text.
Scrape
Extract content from individual pages with precision. Optimized for single-page extraction with CSS selectors and metadata.
Search
Perform search engine queries and automatically crawl the results. Discovery + extraction in one step.
Output
Screenshot
Capture high-quality screenshots of any web page. Full-page or viewport, as base64 or binary.
Transform
Convert raw HTML or PDF into clean markdown, plain text, or sanitized HTML. Process content offline without re-fetching pages from the web.
Links
Collect all links from a website without extracting page content. Optimized for sitemap generation and URL discovery at lower cost.
What every endpoint already does.
Proxy support
Residential proxies with geo-routing to 199+ countries
Streaming
JSONL streaming so you can process results as they arrive
Caching
Built-in HTTP caching with configurable TTL
Webhooks
Async delivery when crawls complete
Rate limiting
Automatic concurrency management
Auth header
Simple Bearer token on every request
Multi-format
JSON, XML, CSV, or JSONL output
Client libraries
SDKs for Python, JavaScript, Rust, Go, and CLI
Skip the REST calls. Use the MCP server instead.
All 9 endpoints (plus browser automation and AI tools) are available as MCP tools. Connect Claude, Cursor, or Windsurf directly.
npx spider-cloud-mcp or mcp.spider.cloudBandwidth and compute, billed by the request.
Spider's figure is the pricing guide's estimate for a mixed workload in smart mode. Heavy pages and residential proxies cost more.
$8 to $15 per 100K pages, pay as you go
Each vendor shows its cheapest plan that covers the volume, at one credit per plain page fetch, from list prices read on September 2, 2026. Every price is month to month. The ratio is against the top of Spider's range.
How spider compares
Monthly cost by volume.
Spider charges bandwidth and compute with no credit multipliers.
| Provider | 10K pages | 100K pages | 1M pages | Multipliers | Credits expire |
|---|---|---|---|---|---|
| Spider | $0.80 to $2 | $8 to $15 | $80 to $150 | None | Never |
| Firecrawl | $99 | $99 | $749 | None | Monthly |
| ScrapingBee | $19 | $49 | $99 | 1 to 75x | Monthly |
| Apify | $0.00 | $19 | $67 | Per CU | Monthly |
| Scrapfly | $30 | $30 | $100 | 1 to 30x | Monthly |
| Bright Data | $15 | $150 | $1,301 | Per product | Monthly |
Spider is the pricing guide's mixed-workload estimate. The others are the cheapest plan that covers the volume at one credit per plain page fetch, from list prices read on September 2, 2026.
Unlimited: one monthly price, sized by concurrency.
Scrape, crawl and links run on /unlimited/scrape, /unlimited/crawl and /unlimited/links with the parameters and API key you already use. You pay for slots, not pages.
Priced by concurrency
You buy how many requests can run at once, from 2 for $40 a month. Each added slot costs less, from $20 down to $13.
An answer, never a queue
With every slot busy you get a 429 right away, with Retry-After. Responses carry X-Concurrency-Limit and X-Concurrency-Active.
Stealth takes one more slot
Send stealth: true and the request uses one extra slot while it runs. It is never refused for that; on a full plan it waits a moment instead.
From signup to first crawl.
A Bearer token, a URL, a curl. That's the first request.