# Plainscrape

> A question-first web extraction API for coding agents. Point it at a URL, ask in plain language, get structured JSON back. No selectors. No XPath. Failed pages are free.

Plainscrape is a metered web API. You submit a URL and a plain-language question; you get structured JSON back. The billable unit is a page fetched, not a request and not a credit. JavaScript-rendered pages are handled for you. New accounts get a one-time 100-page trial (14 days, no card). After that, pages come from one-time packs; nothing resets on the first of the month.

## Product

- Home: https://www.plainscrape.com/
- Docs: https://www.plainscrape.com/docs.html
- MCP setup: https://www.plainscrape.com/mcp.html
- Cursor: https://www.plainscrape.com/integrations/cursor
- Claude Code: https://www.plainscrape.com/integrations/claude-code
- JavaScript-rendered pages: https://www.plainscrape.com/glossary/javascript-rendering
- Without CSS selectors: https://www.plainscrape.com/glossary/without-selectors
- Structured JSON, not HTML: https://www.plainscrape.com/glossary/structured-json
- Pricing: https://www.plainscrape.com/pricing.html
- Why selectors fail: https://www.plainscrape.com/why.html
- Benchmarks: https://www.plainscrape.com/benchmarks.html
- Playground: https://www.plainscrape.com/playground.html

## For agents (MCP)

HTTPS endpoint: `https://www.plainscrape.com/mcp`

Five tools, same verbs as the REST API: `extract`, `extract_page`, `answer`, `get_results`, `account`. Auth is `x-api-key`. There is no keyless MCP.

Works with Cursor, Claude Code, Windsurf, VS Code, and Cline.

Prefer `answer` when you have a homepage URL and a question. Prefer `extract` / `extract_page` when you already have the listing or detail URL.

Completed jobs return a `download_url` when result storage is configured. The live poll body via `get_results` expires after 24 hours; the download link provides durable access beyond that window.

Live-green examples (max_pages 1 unless noted):

- extract https://quotes.toscrape.com/ — every quote with text and author
- extract https://books.toscrape.com/ — every book with title, price, rating and availability
- extract https://news.ycombinator.com/ — every story with title and link (do not require points/author)
- extract https://blog.python.org/ — every post with title, date, and link
- extract_page https://books.toscrape.com/catalogue/a-light-in-the-attic_1000/index.html — title, price, rating, availability (rating may be null; see missing_fields)
- answer https://books.toscrape.com/ — What categories of books are listed, and roughly how many books appear on the first page? (max_pages 2)

## Pricing (true today)

- Trial: 100 pages, 14 days, one-time, no card.
- Pack: $16 for 1,000 pages, one-time, good for 12 months from purchase.
- Volume above 1,000 pages is quoted at enterprise@plainscrape.com — there is no published per-page rate for larger batches.
- Failed and blocked pages are not billed.
- `max_pages` is the cost cap and is clamped server-side at 10. Over-limit requests still run, at 10 pages.
- There is no monthly subscription SKU. Packs do not renew. Unused pack pages do not reset on the 1st.

## What it is not

- Not a bulk-crawl or lead-gen enrichment product.
- Does not return CSV or markdown today (JSON only). Those are on the docs roadmap.
- No SOC 2 report.
- No refine-without-recrawl endpoint.
- `use_proxy: true` is optional (default false) for sites that block datacentre IPs; leave it off otherwise.

## Contact

- hello@plainscrape.com — support
- enterprise@plainscrape.com — volume
- security@plainscrape.com — vulnerabilities
- Legal entity: DataCortex IQ Private Limited