Skip to content

Get the data.
Leave the web to us.

Point Vermin at a URL. It slips past the JavaScript, chews off the cookie banners and navs, and drags back markdown your model can actually use.

Start building Read the docs

Free: 1,000 pages a month. No card.

api.vermin.dev · /v1/scrapereplay
POST /v1/scrape https://shop.example/p/42 formats: markdown, chunks▸ fetch     GET 200 · 38 KB · 41 ms✕ quality   js_required (empty <div id=root>)↑ escalate  render · headless chromium▸ render    200 · hydrated · 612 ms✓ extract   3,412 tokens · 8 chunks · −41% junk tier render  credits_used 1  latency 703 msgot in through the window. left with the data.
try →
91.7%
of pages returned usable content
81.8%
success on bot-protected pages
98.3%
metadata accuracy

Measured on 120 stratified pages, 2026-09-29. Built for PageClerk, where it lifted metadata coverage from 49% to 99%. Our story →

§01Escalation

Starts cheap. Digs deeper.

You don't pick a mode. Vermin tries the front door, checks what came back, and only tunnels down to a browser when the page needs one.

  1. 0 m

    Fetch

    Plain HTTP, then a quality check. Fast, cheap, and enough for most of the web.

    1 credit
  2. −1 m

    Browser

    Empty JavaScript shell? A headless browser renders it and waits for the content.

    same price
  3. bedrock

    Still blocked

    Bot wall, timeout, 404. You get a typed error code and keep your credit.

    0 credits

§02One API

A page, a site, or the whole web.

Scrape, crawl, map, search, extract and brand. One key, one credit balance.

LLM-ready

Your model doesn't need the footer.

Navs, cookie banners and footers go before your model sees them. On a crawl, anything repeated across the site goes too.

  • Home · Products · Pricing · Blog · Log in (removed)
  • 🍪 We value your privacy. Accept all / Manage (removed)
  • # Rat-proof storage bins
  • Our bins are made from 1.2 mm steel with a gasket lid…
  • ## Sizes
  • | Size | Litres | Price |
  • You may also like: Mouse trap deluxe, Cheese (bulk) (removed)
  • © 2026 Example Inc · Terms · Privacy · Sitemap (removed)

kept 0 tokens

removed 0 tokens

Brand

Logo, colours, fonts.

Any domain, one call, cached globally.

vermin.dev

font Bricolage Grotesque · 5 credits

Map

Every URL on a site.

Sitemaps, robots, link graph. Ranked by your query.

0

urls found · 0 duplicates

Chunks

Ready for a vector store.

Each chunk carries its heading path and token count.

  • Rat-proof storage bins318
  • Rat-proof storage bins › Sizes204
  • Rat-proof storage bins › Care156

Search

Don't know the URL?

Search the web, then scrape the results in the same call.

q best rat-proof bins

  1. 1Rat-proof storage bins, testedbinreviews.example/rat-proof
  2. 2Keeping rats out of the shedgarden.example/rats
  3. 3Steel vs plastic binsforum.example/t/8812

5 credits per 10 results

Extract

Typed JSON, not prose.

Give it URLs and a schema. Get fields back, validated.

shop.example/bins/steel-60

{
  "name": "Steel bin 60L",
  "price": 34.5,
  "inStock": true
}

15 credits per URL · failures free

Crawl

Whole sites, streamed.

Sitemap-first discovery, include/exclude globs, live SSE stream or signed webhooks. Respects robots.txt.

event: page /                 fetch      +1event: page /pricing          fetch      +1event: page /app/dashboard    render     +1event: page /blog/launch      fetch      +1event: skip /admin            robots.txt +0event: page /docs/quickstart  fetch      +1

§03Agents & developers

Use it from anywhere.

Tools in Claude, Cursor or any MCP client. A CLI for scripts and agents. Typed SDKs with retries and streaming crawls. Or plain cURL.

Read https://vermin.dev/agent.md and set up Vermin for this project.
Use it whenever you need to read a web page, crawl a site, search the web, pull structured JSON from pages, or get a company's brand.

1 credit is 1 page, browser included. Failed pages are free.

See pricing

§04FAQ

Questions, answered.

What is Vermin?

One API that turns any URL or website into clean, LLM-ready data: markdown, token-counted chunks and metadata, structured JSON validated against your schema, plus brand data like logos, colours and fonts. Scrape a page, crawl a site, map every URL on it, or search the web.

Do I pay for pages that fail?

No. If a page is blocked, times out or returns nothing usable, the response has a typed error code and credits_used is 0.

Does it handle JavaScript-heavy sites?

Yes. Every response is quality-checked. If a plain fetch comes back as an empty JavaScript shell, Vermin escalates to a headless browser automatically. You never pick a mode.

Can my AI agent use it?

Yes. Connect the hosted MCP server at mcp.vermin.dev or run npx -y @vermin/mcp locally. Agents can also read vermin.dev/agent.md to set themselves up.

Release the vermin.

The web is messy. Your data doesn't have to be. Start free, no card.

Start building Read the docs