AutoScout24 Clean Scraper with 75 typed fields, 8 EU domains

AutoScout24 Clean Scraper — 82 typed fields, 8 EU domains

🎁 First 10 records virtually free (~$0.0001 total). ~$0.36 for 100, ~$4 for 1,000. Pay only per delivered record above the trial.

Pull AutoScout24 used-car listings as clean structured JSON across 8 EU domains (DE/FR/IT/ES/NL/AT/BE/EN) — 82 ready-to-query fields per vehicle. VIN, GPS coordinates, VAT recoverability, accident history, TÜV inspection date, dealer hero images and opening hours — fields competing scrapers don't deliver. Cross-locale enums normalized to EN-stable values with raw locale preserved.

Built for data engineers shipping dealer pipelines. Also powers market analytics dashboards and AI agents (MCP-native).

✅ Production-hardened — 100+ runs, 479 tests, 5 review-loop cycles completed 2026-05-20. ✅ Idempotent — stable vehicleId UUID dedup survives Apify host migrations. ✅ Soft-block aware — Cloudflare / Captcha / WAF challenges trigger automatic session rotation. ✅ Cost-correct — failed records never charge; transient billing outages auto-recovered (4-attempt exponential backoff).

---

Why this scraper

🎯 82 ready-to-query fields per listing. Integers are integers (mileageKm: 58500), dates are ISO 8601 (firstRegistration: "2019-06-01"), equipment is structured arrays — never a "58,500 km" string to parse.

📍 The columns B2B teams actually filter on. VIN (when published by dealer), GPS coordinates (100% populated), VAT recoverability, vehicle history (accidents, previous owners, full service history, next TÜV inspection). Standard in this actor, absent everywhere else.

🛡️ Pay-per-event billing. Idempotent + actively maintained. You only pay for successful records pushed to your dataset — not for proxy retries, blocked fetches, or partial run failures upstream of push. Note: records pushed before a mid-run crash ARE billed (standard Apify PPE semantics). Stable vehicleId UUID dedup, schema versioning via _meta.parserVersion, < 24h support response.

---

Why this Actor beats $4 alternatives

Most AS24 scrapers on the Store ship locale-leaked strings you have to clean up downstream. Side-by-side, what you actually get per record:

What you get	Typical AS24 scraper	This actor
VIN	absent	identifier.vin: "WAUZZZF49KN012345" (when published by dealer)
Mileage	"58,500 km" string	58500 integer
Price	"€ 119,900" string	{amount: 119900, currency: "EUR", vatDeductible: false, vatRate: null, netAmount: null}
Transmission	"Schaltgetriebe" (DE locale)	"Manual" + transmissionOriginal: "Schaltgetriebe"
GPS	absent	{latitude: 50.96258, longitude: 7.18602}
Equipment	CSV string "ABS,ESP,Klimaautomatik,..."	Structured arrays per category (4 categories, typed entries)
Vehicle history	absent	{hadAccident: false, previousOwners: 3, fullServiceHistory: true, hsnTsn: "0583/AJZ"}
Dealer enrichment (v1.1)	name + phone string	seller.{heroImage, whatsappNumber, homepageUrl, contactPerson:{name,position,languages[]}, openingHoursByDay:{monday:{open:"09:00",close:"18:00"}, sunday:{closed:true}, …}, googleRatings} — extracted from the same listing payload, zero extra fetches
Pipeline-ready	parse strings manually	direct JOIN on integers/ISO dates/enums
Cross-domain consistent	locale-leaked	EN enums + original preserved in *Original sister fields
Migration-resilient	trial window re-opens / counters reset	_record_count persisted to KVS + Event.MIGRATING handler → cap math survives Apify host migrations
Soft-block aware	silent partial dataset on Cloudflare challenge	SoftBlockSessionError detection (CF / Captcha / WAF sniff) → automatic session rotation
Numeric drift tolerance	parser crashes on AS24 schema drift	Locale-aware weight regex (DE/EN/IT/ES/NL/FR) + safe coercion across 7 engine fields (CO₂, range, consumption, power)
Cost-correct PPE	charged for records lost to race conditions	Abort race fix + charge retry (1s/3s/9s exp backoff) → no pushed-not-charged or charged-not-pushed records

---

Try it in 60 seconds

1️⃣ Apify Console (no code)

Click Try for free above. Paste any AS24 search URL. Hit Save & Start. First records flow in within ~10 seconds.

Example URL to paste:

https://www.autoscout24.com/lst/porsche/991?sort=age&desc=1

2️⃣ AI Agent (MCP-native)

Plug directly into Claude Desktop, Cursor, or any MCP client. The agent calls the scraper with natural language — no glue code:

https://mcp.apify.com?tools=hardy_ice-owner/autoscout24-clean-scraper-py

3️⃣ REST API

curl "https://api.apify.com/v2/acts/hardy_ice-owner~autoscout24-clean-scraper-py/run-sync-get-dataset-items?token=YOUR_TOKEN" \
  -X POST -H "Content-Type: application/json" \
  -d '{"urls": ["https://www.autoscout24.de/lst/audi/a6"], "maxRecords": 50}'

4️⃣ Python SDK

from apify_client import ApifyClient

client = ApifyClient("YOUR_TOKEN")
run = client.actor("hardy_ice-owner/autoscout24-clean-scraper-py").call(
    run_input={"urls": ["https://www.autoscout24.de/lst/audi/a6"], "maxRecords": 50}
)
for vehicle in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(vehicle["vehicle"]["make"], vehicle["vehicle"]["model"], vehicle["price"]["amount"])

---

Real output sample

One record, the differentiators visible at a glance:

{
  "id": "d6a3056e-267d-45e6-a160-4777491ddcde",
  "identifier": { "vin": "WP0ZZZ99ZKS123456", "offerReference": "P-2024-1287" },
  "vehicle": {
    "make": "Porsche", "model": "991",
    "modelVersion": "911 Carrera 4 GTS",
    "engine": { "powerKw": 331, "powerHp": 450, "cylinders": 6, "gears": 7 }
  },
  "price": { "amount": 119900, "currency": "EUR", "vatDeductible": false },
  "mileageKm": 58500,
  "firstRegistration": "2019-06-01",
  "nextSafetyInspection": "2026-06",
  "location": { "latitude": 50.96258, "longitude": 7.18602, "city": "Bergisch Gladbach" },
  "history": { "previousOwners": 3, "fullServiceHistory": true, "hsnTsn": "0583/AJZ" },
  "fuel": { "co2EfficiencyClass": "Green", "emissionStandardLabel": "EURO6" },
  "_meta": { "parserVersion": "ndata-v2", "fetchMethod": "http" }
}

…plus 60+ more fields (full schema: equipment arrays per category, HD/mid images, dealer info, ratings, description HTML+text, etc.).

---

Pricing — try (virtually) free, pay only above the trial

🎁 First 10 records every run cost ~$0.0001 total (1/100 cent — Apify Store min event price). No card needed for a real evaluation. Convert when ready.

$0.004 per delivered record above the trial. No subscription. No minimum. Failed records never charge.

Run size	Trial (1-10)	Paid (after 10)	Total
10 vehicles	10 × $0.00001	—	$0.0001
100 vehicles	10 × $0.00001	90 × $0.004	$0.36
1,000 vehicles	10 × $0.00001	990 × $0.004	$3.96
5,000 vehicles	10 × $0.00001	4,990 × $0.004	$19.96
50,000 vehicles	10 × $0.00001	49,990 × $0.004	$199.96

Volume 50k+/mois : offre sur mesure via Apify Console messaging on the actor page.

Proxy data metered separately by Apify (free tier covers small runs; large jobs may need RESIDENTIAL proxy).

Premium options (priced higher)

Two opt-in flags trigger premium per-record pricing:

Option	Price/record	When to use
Default (HTTP)	$0.004	Standard scrape — works on 99% of AS24 pages
forcePlaywright: true	$0.012 (3×)	Browser-rendered fallback — enable only if HTTP gets blocked at high volume
includeRawData: true	$0.005 (+25%)	Adds raw __NEXT_DATA__ JSON blob (5-10× output size) — for debugging, custom field extraction, or downstream raw shipping

Stacking applies: forcePlaywright + includeRawData → $0.012 (Playwright cost dominates).

Premium options ALSO honor the 10-record virtual trial ($0.00001 each).

The Console UI labels both options with ⚠️ PREMIUM so you see the cost impact before running.

---

Standard vs Full — which one?

Both tiers ship the same 82 typed fields. Full appends the raw __NEXT_DATA__ payload under _raw.* for everything we don't normalize.

Standard ($0.004)  Full ($0.005)
                       =================  =============
82 typed fields            ✓ all 82            ✓ all 82
+ raw __NEXT_DATA__        —                   ✓ (~150 extra fields)
Record size                ~5 KB               ~38 KB (6.8×)
Bandwidth (1000 rec)       5 MB                38 MB

Use Standard if you need:

• Dealer CRM ingestion (clean fields → DB columns)
• Analytics dashboards (price/km curves, regional supply)
• Cross-source dedup via UUID
• AI agent queries via MCP
• 95% of B2B pipelines

Use Full if you need:

• _raw.dpvStatistics — page-view & favorite counts (lead scoring)
• _raw.financingAndInsurance — AS24-native leasing/finance offers per record
• _raw.prices.public / _raw.prices.dealer — split B2C vs B2B price evaluation
• _raw.adTier / appliedAdTier — dealer ad spend visibility
• _raw.adTargetingString — AS24 segmentation cohort
• Custom field extraction (you need a field we don't parse)
• Forward raw payload to your data lake
• Debug parser regressions

---

Battle-tested — 5 hardening cycles, 479 tests, 0 regressions

Most AS24 scrapers on the Store are write-once / hope-it-still-works. This actor went through 5 successive review-loop cycles during the week of 2026-05-19/20 (loop 1 → loop 5), each one fixing a concrete production failure mode before publish. +105 new tests, +28% coverage, zero regression across the cycle. Numbers are the headline; the why-it-matters is below.

#	Loop	What was fixed	Why it matters to your pipeline
1	Performance	Parallel HEAD preflight (50s → 5s), vehicle cache (−7 redundant lookups/record), regex HTML strip (10× faster than BeautifulSoup), KVS handle reuse	+18–25% throughput on large runs; cap-math fires sooner so you stop closer to maxRecords
2a	Cost-correctness (PPE)	Abort race fix (no more pushed-not-charged), record_pushed lifted out of finally (no counter inflation on retry), defensive chargeLimit detection, charge retry with 1s/3s/9s exponential backoff	Recaptures ~$0.30–0.40 per 1k records previously lost to billing transients. No double-charges, no silent missed-charges.
2b	Migration safety	_record_count persisted to KVS + Event.MIGRATING handler registered	Trial window + cap math now survive Apify host migrations. Pre-fix: a migration mid-run could silently re-open the $0 trial window on the new host, breaking pricing invariants.
3	Parser robustness (ndata-v2.1)	Media crash guard (None URL filter), CO₂ / electric-range / consumption coercion via _safe_int / _safe_float, evaluationCategory null union, weight regex locale-aware (DE thousands-dot, FR space separator, EN/IT/ES/NL plain int), numeric type-drift protection across 7 engine fields	Drift-tolerant: when AS24 ships a stray "45.5" where they used to ship 45, the record validates instead of crashing. Worth its weight in any long-running schedule.
4	Fetch reliability (v1.1.3)	SoftBlockSessionError subclass + Cloudflare / Captcha / WAF challenge sniff, detailUrlsOnly preflight, HEAD UA spoofed to Chrome 126, deterministic 404/410/451-only filter (no false-positive on 403 / 429 / 5xx), startswith prefix bug fix, dead-config cleanup	Soft-blocked requests rotate the session instead of polluting the dataset with empty records. Preflight only filters genuinely-gone URLs, not transient errors.
5	Sign-off	479 tests deterministic, ruff/mypy clean, 5 cross-loop integration checks PASS	Ship signal — no flake, no skip, no # type: ignore churn

Test count progression across the cycle: 374 → 421 → 432 → 465 → 479 + 1 xfail (intentional, schema-evolution gate). All 5 commits land on main; Apify webhook rebuilds the published actor on push.

What this means for a buyer: when AS24 changes their HTML, ships a new locale string, or returns a Cloudflare interstitial mid-run, you don't get paged at 3am — you get a clean record or a clearly-tagged retry. That's the moat.

---

Use cases

Dealer inventory monitoring — $6/day

Track 50 competitor dealers hourly. ~1,500 vehicles/day. Apify Schedules + webhook → your CRM. Cheaper than 1h of analyst time.

Market analytics — $20/week

5,000 BMW 3-Series scraped weekly. Model price erosion, mileage curves, regional supply. CSV/Excel export native.

AI agents (MCP-native) — pay per query

Connect via mcp.apify.com to Claude, Cursor, or any MCP client. Agent asks "2023 Tesla Model 3 under €40k near Berlin"; gets structured JSON; answers naturally. ~$0.04 per 10-vehicle query.

---

Operational notes

• 8 EU domains supported: .de, .com, .fr, .it, .nl, .at, .be, .es. Cross-domain output normalized via EN-stable enums (bodyColor, vehicleType, emissionStandardLabel).
• Cross-domain enums normalized: transmission, drive train, body type, upholstery, color, paint type, original market — all mapped to EN-stable values from DE/FR/IT/NL/AT/BE/ES locales. Raw locale preserved in *Original sister fields.
• AutoScout24 caps any search at ~400 results (20 pages × 20 listings). For bigger queries, split by region/year/price-range and pass multiple urls. The actor dedups across them.
• Performance: 1,000 records in ~4 min; 5k in ~25 min; 50k in ~4 h. Memory tier 4 GB default; smart-capped to prevent OOM.
• Idempotency: intra-run dedup by AS24 UUID is 100% — no duplicate vehicle within a single run. Cross-run overlap varies (~80–85% on sort=age/sort=price URLs over a few minutes) due to AS24 listing volatility (new entries, sold items removed, paginator non-determinism). Stable id field allows union dedup downstream across N runs.
• maxRecords overshoot bound: actor may overshoot maxRecords by ~10% due to in-flight handlers when the stop gate fires (e.g. request maxRecords=1000, receive up to ~1100 — billed accordingly). Set the Apify PPE charge limit to your hard budget cap; the actor stops cleanly when either bound is hit.
• Schema stability: _meta.parserVersion = "ndata-v2". Bumped only on breaking parser changes — your pipelines can gate on this field.

---

FAQ

Q: Is scraping AS24 legal in the EU? A: Public listings only, no authentication bypass, no PII beyond what AS24 publishes on its public marketplace. We respect AS24 rate limits.

Q: Can I skip listing crawl and hit detail URLs directly? A: Yes. Use detailUrlsOnly: ["https://www.autoscout24.de/offers/audi-a6-...", ...] instead of urls.

Q: How do I scrape beyond the 400-result AS24 cap? A: Provide multiple urls — one per filter slice (year, region, price band). The actor dedups by vehicleId across them.

Q: What if AS24 blocks the scrape? A: Set forcePlaywright: true to render via headless Chrome. Slower (~4×) but bypasses most blocks. Auto-fallback in V1.5.

Q: Can I get CSV or Excel instead of JSON? A: Yes — at download time. Apify Console (Export button) or REST GET /datasets/{id}/items?format=csv|xlsx. The actor always writes JSON.

Q: How fresh is the data? A: Each record carries _meta.scrapedAt (when we fetched) and createdAt (when AS24 published the listing). Re-run on-demand or via Apify Schedules.

Q: Support response time? A: < 24h on weekdays. Issues, feature requests, and private deals via Apify Console messaging on the actor page.

---

Full feature comparison

vs typical AS24 scrapers on Apify Store:

Dimension	Typical	This actor
Field count	~30 (mostly strings)	82 typed
VIN	absent	present when published by dealer
Billing model	flat (failed runs charged)	PPE (only successful records charged)
Cost-correct retry	no	4-attempt exponential backoff on billing transients
GPS coordinates	absent	lat + lng 100%
VAT info	absent	vatDeductible + vatRate
HD images	120×90 thumbnails	1280×960 + 640×480 mid
Vehicle history	absent	accidents, owners, service, TÜV
Equipment	CSV mono-block string	structured arrays
Dealer enrichment	name + phone	heroImage, WhatsApp, opening hours, contact person, Google ratings (zero extra fetches)
Schema versioning	none	_meta.parserVersion
MCP-ready	no	yes
Soft-block aware	no	CF / Captcha / WAF sniff → session rotation
Idempotent across Apify host migration	no	yes — Event.MIGRATING handler + persisted counter
Drift-tolerant types	parser crashes on schema drift	safe int/float coercion across 7 engine fields
Idempotency	unspecified	documented (UUID dedup)
Source URL	inconsistent	canonical
Test suite	none / undocumented	479 tests + 5 review-loop cycles
Maintenance	varies	< 24h support, active releases

---

Output schema

_meta.parserVersion = "ndata-v2". 15 logical groups × 82 fields (v1.1 adds 7 dealer-enrichment fields under seller). Top-level: id · sourceUrl · sourceDomain · identifier · price · vehicle (+ nested engine) · mileageKm · firstRegistration · nextSafetyInspection · createdAt · fuel · body · history · location · seller · ratings · equipment (4 categories) · media (HD + mid) · description · status · _meta.

---

Agent-ready (x402, USDC on Base)

For AI agents paying via HTTP 402 micropayments (USDC, no Apify account):

# 1. Install Apify MCP client
npm install -g @apify/mcpc

# 2. Connect to Apify MCP server with x402 payments enabled
mcpc connect "mcp.apify.com?payment=x402" @apify --x402

# 3. Call this actor — mcpc handles signing + retries automatically
#    First 10 records virtually free ($0.00001 each); above 10 = $0.004 USDC/vehicle.

Min $1 USDC prepaid balance on Base mainnet; subsequent calls draw down. Compatible with Coinbase Wallet, Privy, any Base-compatible wallet. PPE billing model required (usesStandbyMode: false, ✓ here).

See Apify x402 docs for wallet setup options.

---

Changelog

v1.1.3 (2026-05-20) — Fetch reliability hardening (loop 4). SoftBlockSessionError subclass + Cloudflare / Captcha / WAF challenge sniff → automatic session rotation instead of polluting the dataset with empty records. detailUrlsOnly preflight added. HEAD requests spoof Chrome 126 UA to dodge bot-detection on the preflight tier. Preflight filter tightened to 404 / 410 / 451 only — no false-positive elimination on transient 403 / 429 / 5xx. startswith prefix bug fixed; dead config keys cleaned up. +14 tests (→ 479).

v1.1.2 (2026-05-20) — Parser robustness sweep (loop 3, ndata-v2.1). Media-list crash guard (None URL filter). CO₂ / electricRange / consumption now coerced via _safe_int / _safe_float — no more parser crash on AS24 numeric drift. evaluationCategory union widened to nullable. Weight regex now locale-aware: DE thousands-dot (1.450 kg), FR space separator (1 450 kg), EN/IT/ES/NL plain int (1450 kg) — drift-tolerant across 7 engine fields. +33 tests (→ 465).

v1.1.1 (2026-05-20) — Cost-correctness + migration safety (loops 2a/2b). Abort race fix: no more pushed-not-charged records when a run is aborted mid-flight. record_pushed lifted out of finally block so retried records don't inflate the counter. Defensive chargeLimit detection. Charge retry with 1s/3s/9s exponential backoff → transient billing-API outages recover automatically. _record_count persisted to KVS + Event.MIGRATING handler registered → trial window and cap math now survive Apify host migrations. +11 tests (→ 432).

v1.1.0 (2026-05-20) — Dealer enrichment + perf wins (loop 1). 7 new nullable seller.* fields surfaced from the listing payload (zero extra fetches) — heroImage, heroImageInterior, whatsappNumber, homepageUrl, contactPerson (name/position/image/languages), googleRatings, openingHoursByDay (weekday-keyed {open, close} or {closed: true}). Closes the dealer-data gap vs $1.29/1K competitors who upsell on the same fields. Field count 75 → 82, all backward-compatible. Parallel HEAD preflight (50s → 5s), dict-wrap drop, vehicle-data cache (−7 lookups/record), regex HTML strip (10× faster than BeautifulSoup), KVS handle reuse → +18–25% throughput. Virtual free trial: first 10 records of every run billed at $0.00001 via new vehicle-trial PPE event. +47 tests (→ 421).

v1.0.5 (2026-05-19) — Filter loop gaps closed: Andorre, Partes en cuero (ES upholstery), MPV (NL bodyType) — 100% × 8 EU locales × 7 fields normalization coverage achieved.

v1.0.1 (2026-05-18) — Pre-publish hardening: abort handler stops PPE billing + crawler on user-aborted runs (was billing up to ~20 records post-abort), description HTML capped at 100 KB before parse, telemetry counter cardinality bounded at 1000 unique keys, URL host allowlist (8 AS24 domains) to block SSRF, KVS dedup persist cadence 50 → 500 (cuts O(n²) write amplification on 50k runs), apify SDK pinned to >=3.4,<3.5 for reproducible deploys.

v1.0.0 (2026-05-17) — Initial release. 75-field schema with cross-domain enum normalization (ndata-v2) — 7 fields mapped to EN-stable values, raw locale preserved in *Original sister fields. MCP + x402 support, PPE pricing, listing + detail crawling, vehicleId dedup, Playwright fallback, 8 EU domains. (Bumped to 82 fields in v1.1.0 — see above.)

---

Support

• Issues / feature requests / custom fields / private deals: Apify Console messaging on the actor page
• Response time: < 24h on weekdays