The scraping API that doesn't rot
haul is a self-hosted scraping and extraction API. One endpoint, four escalating tiers — from a plain fetch to a managed unblocker — so you get the page whether it's static HTML or behind Cloudflare. No per-page credits, no silent caps.
$ curl https://api.haul.sh/v1/scrape \
-H "Authorization: Bearer $HAUL_KEY" \
-d '{"url":"https://stripe.com/pricing","formats":["markdown"]}'
{
"status": 200,
"tier_used": "fetch",
"cache_hit": false,
"markdown": "# Pricing\n\nStart building today..."
}One endpoint. Four tiers.
You never choose a tier. haul starts with the cheapest fetch and escalates only when a site fights back — so easy pages stay fast and hard pages still come back.
Tier 1 · Plain fetch
A fast HTTP GET with a real browser fingerprint. Handles the majority of static and server-rendered marketing, docs, and pricing pages — for free.
I want to generate an image of two people, fighting outside a bar. They fight to the core. Once they're done, they sit down and drink beer.
Certainly, I'm generating this picture for you in a while. BTW are you talking about THAT movie?
I don't know what you're talking about.
Are you sure?
Yes, I'm sure. But if you're generating that scene, make sure the fighters have clown shoes and rubber chickens instead of fists!
Affirmative, here's your image.
Tier 2 · Headless browser
Renders JS-heavy pages and SPAs in real Chromium when a plain fetch comes back as an empty shell.
Tier 3 · Proxy pool
Routes through rotating residential proxies when a site blocks datacenter IPs — without you changing a single line of code.
Add LLM
Add
Groq LLM
23rd March
OpenAI GPT0
21st March
Stable DIffusion
3rd May
Llama 2
1st April
Claude 200k
2nd June
Tier 4 · Managed unblocker
Hands the hostile few to a managed unblocker as a last resort. You never pick a tier — the gateway escalates automatically on real block signals.
Cheapest-first. Fetch → browser → proxy → unblocker, escalating only on real bot-block signals.
Turn any page into JSON that matches your schema — LLM extraction with validation and a corrective retry.
A cross-key Redis cache means a URL is scraped once and reused everywhere. Fast, and cheap.
Per-host concurrency, a min-delay between hits, and robots.txt respected out of the box.
Async jobs: scrape a list of URLs or BFS-crawl a whole site, then poll one job id.
Watch a URL and get a webhook only when the content actually changes — not on every poll.
Prometheus metrics and per-key usage counters. No more pipelines that rot in silence.
Runs on your own Railway or Docker. No per-page credits, no vendor cap to slam into.
Stop watching your data pipeline rot.
haul is in private beta. Request a key, point your scrapers at one endpoint, and get the freshness and visibility you never had on a metered API.
haul