DomsectDomsect
PUBLIC WEB DATA INFRASTRUCTURE

Compiling Order from Web Chaos

Turn millions of raw, noisy DOM nodes into clean structured JSON.

High-throughput web extraction built for modern AI agents and data pipelines.

POSThttps://api.domsect.io/v1/extract
Schema-validated outputStarting at $0.018/page10 free credits · up to 500 via Sandbox
01 · Playground

Experience Pure Signal

Prune 97%+ of DOM noise into structured JSON in milliseconds.

Target Site Coverage Note

DomSect Universal Extract API is optimized for standard, publicly accessible web architectures (e-commerce catalogs, technical documentation, news, public directories). Sites guarded by aggressive anti-bot mitigations (e.g., Amazon, Zillow, LinkedIn) require tailored scraper configurations.

Need custom scraper engineering? Active subscribers can request tailored crawler scripts via support.

Requires: 1 Credit (Iris)

Request preview
{
  "url": "https://poshmark.com/listing/HOT-BUYAuthentic-Burberry-Wallet-with-box-and-dust-bag-68c9b442c1c79e08d59952bc",
  "schema_definition": {
    "title": "str",
    "price": "str",
    "condition": "str",
    "seller_name": "str",
    "specifications": "dict"
  },
  "engine": "iris",
  "stealth": false
}

Demo runs on a shared sandbox key and is rate-limited — get your own key below.

response.json

$ domsect extract
Output will appear here — structured, typed, ready to pipe.

02 · Benchmarks

Measured, not marketed.

Every number below is something we actually ran. No projections, no SLAs disguised as benchmarks.

97%+
HTML denoise rate

Measured: ~2M chars of raw HTML pruned to ~50K before extraction. Tokens are spent on content, not markup.

5/5
Schema-valid JSON, every run

Every model in our benchmark returned schema-valid JSON on the reference page. A measurement on that page, not a guarantee for every page.

$0.018
Flat per page

Predictable, transparent page-based credits. Effective $0.018 on Starter, scales down to $0.011 on Business. Up to 80% less than feeding raw pages into frontier LLMs.

~2.0–2.6s
Extraction time (measured)

Observed on the reference page. A measurement, not a guarantee — network and page weight dominate.

Up to 80% less than feeding raw pages into frontier LLMs

Measured, not estimated: an unpruned product page feeds ~30,000 input tokens into the model (our benchmark) — $0.06–$0.075/page in input tokens alone on Claude Sonnet / GPT-4o at current list prices, output billed extra. Our heuristic pre-pruning strips 97%+ of markup noise before inference, so you pay a flat $0.018/page, all-in.

Billing, plainly

Iris costs 1 credit per successful page, Athena 2, Navajo (100% US-Soil Dedicated) 3. Failed fetches, 404s and timeouts cost 0 — you never pay for our misses. Engine internals are server-side and automatic; there is no model parameter to fiddle with.

Compliance boundary

URL-only, public data only — no logins, no paywalls, no private documents. Sensitive hits are blocked at the gateway at zero charge. See the ToS and AUP.

03 · Pricing

Starting at $0.018/page. Zero on failure.

Credits, not tokens. From 1 credit per successful page — 404s and blocks cost 0. Up to 80% less than feeding raw pages into frontier LLMs. Sign in with GitHub for 10 free credits, no card.

See full pricing →
$29/mo
1,600 pages · Effective $0.018/page
$79/mo
5,500 pages · Effective $0.014/page
$199/mo
17,000 pages · Effective $0.011/page

Domain investor? The Domain Screener plans are a separate monthly subscription — different product, different billing.

04 · Quickstart

Copy. Paste. Extract.

01

Get a key

Sign in with GitHub for 10 free credits, or apply for the 500-credit sandbox. Swap in your own key where the samples show *********.

02

Make the call

POST a URL plus a field→type schema. Routing, denoising and model selection all happen server-side — there is no model parameter.

03

Get JSON per schema

The response carries data shaped exactly like your schema on successful runs, plus latency_ms and token usage for observability.

import requests

r = requests.post(
    "https://api.domsect.io/v1/extract",
    headers={"Authorization": "Bearer *********"},
    json={
        "url": "https://www.ebay.com/itm/267768858579",
        "schema_definition": {"title": "string", "price": "string"},
        "engine": "iris",  # "athena" for complex DOM, "navajo" for 100% US-Soil (3 credits/page)
        "stealth": False,
    },
    timeout=60,
)
print(r.json()["data"])
05 · Sandbox access

Start free, scale when ready

Two ways in — pick the one that matches your evaluation.

Layer 1 · Instant self-serve
10 Credits · free
  • · Sign in with GitHub (30+ days old), key is live in 30 seconds
  • · Zero friction — run Hello World immediately
  • · Built for developers who want to feel the API first
Sign in with GitHub →
Layer 2 · Enterprise sandbox review
500 Credits · application
  • · For batch pipelines & schema-consistency evaluation (50–100 pages)
  • · Tell us your company, target sites and monthly volume
  • · Manual review — approved keys are topped up directly
Apply for Sandbox →

No card required for either layer. Layer 2 applications are reviewed by a human — we only use your details to evaluate the sandbox request.