Skip to content

Getting started

Quickstart

From zero to your first scraped page in under two minutes.

Vermin turns any URL, site or search query into clean, LLM-ready data. One key, one base URL (https://api.vermin.dev), six endpoints. Every response tells you exactly what it cost, and anything that fails costs nothing.

1. Grab an API key#

Create a free account (1,000 credits a month, plus 2,000 once your email is verified), then open Dashboard → API keys and create a key. It looks like vmn_live_… and is shown once, so paste it somewhere safe:

shell
export VERMIN_API_KEY="vmn_live_..."

2. Scrape a page#

Pick your weapon. The tabs remember your choice across the docs.

curl -X POST "https://api.vermin.dev/v1/scrape" \
  -H "Authorization: Bearer $VERMIN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "url": "https://example.com", "formats": ["markdown", "metadata"] }'

SDK users: install first (the full list is on the SDKs page).

npm install @vermin/sdk

3. Read the response#

200 OK
{
  "success": true,
  "data": {
    "url": "https://example.com",
    "markdown": "# Example Domain\n\nThis domain is for use in illustrative examples in documents...",
    "metadata": { "title": "Example Domain", "language": "en", "statusCode": 200 },
    "tier": "fetch",
    "cached": false,
    "quality": { "score": 0.97, "issues": [], "escalations": [], "reason": "Plain fetch returned complete content." }
  },
  "credits_used": 1,
  "latency_ms": 184,
  "request_id": "req_01J9V3KX7Q"
}

Three fields worth knowing from day one:

  • tier: which pipeline produced the page. fetch is a plain HTTP fetch, render a headless browser, premium a residential proxy. With the default tier: "auto" we only escalate when the page needs it, and you pay for the tier that worked.
  • quality: why content might be thin (js_required, bot_wall, cookie_wall…) and what the pipeline did about it. No more guessing why a page came back empty.
  • credits_used: what this call cost. Failed requests always report 0.

That's it. You're in.

Where next#

  • Crawl a whole site, streamed live or pushed to your webhooks.
  • Extract typed JSON with a schema. Valid every time, or free.
  • Point an agent at the MCP server.
  • Browse the full API reference, generated from our OpenAPI spec.