realestate.com.au Property Data Extraction

Purpose

Given a realestate.com.au property URL — or a street address that can be resolved to one — return structured property data including the full address, geographic coordinates, listing status, current price (or price range / sold price), key statistics (beds, baths, parking, land size, floor area, property type), the agent + agency, and the property's historical sale prices. Read-only — never submits enquiries, never starts a contact form, never signs in. The same skill handles for-sale, sold, and off-market addresses.

When to Use

• Building a property research dataset (one-shot or scheduled) — e.g. compare current listings to historical sale prices in a suburb.
• Enriching a CRM or buyers' brief with REA-sourced beds/baths/land-size + agent contact.
• Computing capital-gain analytics from historical_sales for a single dwelling.
• Looking up a non-listed address to confirm beds/baths/land before contacting the owner ("off market" path).

Workflow

The richest single page is the Property Pages address profile at /property/{slug}/. It carries the Property history (every recorded sale with date + price + agency), property features (beds/baths/parking/land-size/property-type), the lat/lon coordinates (embedded in the Google Static-Maps URL), and — when the address has an active listing — a deep-link to the /property-{type}-... listing detail page where agent contacts + current asking price + description live. The recommended flow is to land on Property Pages first, then conditionally fetch the listing page if a current price/agent is needed.

A bare cloud session works for the page-rendering surface; PerimeterX/Kasada returns 429 on direct HTTP fetches but does not challenge the headed browser path. Stealth (--verified) and residential proxies (--proxies) are not strictly required for the rendering path on the addresses tested, but keep them on by default — Kasada's bot-score tracker is opaque and some IP ranges hit a captcha interstitial. The fetch + GraphQL paths are confirmed blocked (see Gotchas).

1. Create a stealth session (proxies optional but safer)

sid=$(browse cloud sessions create --keep-alive --verified --proxies \
  | node -e "let s='';process.stdin.on('data',c=>s+=c).on('end',()=>process.stdout.write(JSON.parse(s).id))")
export BROWSE_SESSION="$sid"

2. Resolve to a Property Pages slug

You have one of three inputs:

(a) A listing URL like https://www.realestate.com.au/property-house-vic-richmond-151242968:

• Open the listing first. The breadcrumb (item 5) is the street address. Build the Property Pages slug from it: {number}-{street-name}-{street-suffix-abbr}-{suburb-lower}-{state-lower}-{postcode}. Street suffixes are abbreviated: Street→st, Road→rd, Avenue→av, Parade→pde, Terrace→tce, Court→ct, Drive→dr, Place→pl, Lane→la, Crescent→cr, Highway→hwy. For apartments/units, prefix unit-{unitNumber}-: 5/19 River Street → unit-5-19-river-st-richmond-vic-3121.
• Or skip slug derivation entirely: the listing page itself carries address + beds/baths/parking/land-size/price/agent. Use Property Pages only when historical sales or lat/lon are required.

(b) A sold URL like https://www.realestate.com.au/sold/property-apartment-vic-richmond-150805280:

• Same shape as a listing URL but the price strip reads $815,000 Sold on 02 May 2026, and a [Property history](https://www.realestate.com.au/property/{slug}/) link is rendered just under the price — follow that link rather than rebuilding the slug.

(c) A bare street address (no URL):

• Construct /property/{slug}/ directly using the same slugging rules above. If the slug is wrong, REA serves the closest match in a "Did you mean?" header — pivot to the proposed slug.

3. Open the Property Pages URL

browse open "https://www.realestate.com.au/property/6-stawell-st-richmond-vic-3121/" --remote
browse wait load --remote   # may time out at 30s — non-fatal, the page renders in parallel
browse wait timeout 4000 --remote
browse get markdown body --remote > /tmp/page.md

Important: browse wait load frequently times out at 30s because REA keeps an analytics socket open after DOM-ready. The error is non-fatal — the page IS fully rendered; just follow with a 4s timeout-wait then snapshot.

4. Extract fields from the rendered markdown

The markdown body is structured and predictable. Use these anchors:

Parse the price-range display string with ^\$([\d,]+)\s*-\s*\$([\d,]+)$ for price_low/price_high; treat Contact Agent, Expressions of Interest, Auction, $XXXm shorthand, and single-number displays as price_display only (don't fabricate a range).

5. If a current asking price + agent contact is required, also open the listing page

When Property Pages shows a View listing link, follow it to grab:

• price_display and indicative_price — listing pages often show "Contact Agent" as the headline plus an Indicative price: $X,XXX,XXX - $Y,YYY,YYY sub-line (regulated in VIC by the Statement of Information).
• agent, agency, agent_phone (truncated — REA masks the last 3 digits unless you click "Call", which we don't do).
• Full prose description and AI-generated property_highlights (3 bullets).
• inspection_times, auction_time when present.

For sold listings (/sold/property-...), the same template renders but the price strip becomes ${sale_price} Sold on {date}.

6. Geocoding fallback

If the static-map regex doesn't match (rare, but possible when REA serves a placeholder map), open the listing page — the "Map" embed there carries the same markers=icon%7C{lat}%2C{lon} pattern. As a last resort, the address itself plus the postcode is enough to geocode externally; REA never publishes lat/lon as numeric DOM attributes — it's only embedded in the static-map URL.

7. Release the session

browse cloud sessions update "$sid" --status REQUEST_RELEASE

Site-Specific Gotchas

• Direct HTTP fetch is blocked. browse cloud fetch <listing-url> --proxies returns 429 + Kasada PerimeterX cookies (x-kpsdk-ct, KP_UIDz). The full browser path is the only reliable surface. Don't waste cycles trying to bypass Kasada via raw HTTP, persisted GraphQL queries, or REST endpoints under lexa.realestate.com.au / services.realestate.com.au — they're whitelisted to the mobile app's signed tokens.
• browse wait load times out at 30s on most listing/property pages. REA keeps a long-lived analytics WebSocket open, so the load event never fires. The error is non-fatal — wait timeout 4000 after the open is enough; DOM is ready before wait load would fire anyway.
• A bare session works but margins are thin. Both --verified --proxies and a bare session rendered the same listing/property pages on test runs. However, Kasada is on the perimeter, so traffic spikes / repeated IDs from the same IP push you into a captcha interstitial. Default to --verified --proxies for production; only drop stealth flags if cost matters and the call rate is low (< 10/hr).
• Listing URL structure has TWO formats. /property-{type}-{state}-{suburb}-{id} is the for-sale path; /sold/property-{type}-{state}-{suburb}-{id} is the sold path. Same template but the "Property history" link only appears on the sold variant. The for-sale variant requires a separate Property Pages lookup for historical sales.
• Lat/lon is never published as DOM data — only as Google Static-Maps URL query params. Pattern: markers=icon%3A...%7C{lat}%2C{lon} (URL-encoded | and ,). Regex %7C(-?[0-9.]+)%2C(-?[0-9.]+). Present on Property Pages and the listing's "Map" embed; absent on off-market pages where the dwelling has no street-view coverage.
• window.__INITIAL_STATE__ / __NEXT_DATA__ / __APOLLO_STATE__ are NOT present. REA's frontend is a custom "Argonaut" SPA with no hydration blob exposed on window. Don't try to browse eval for one — extract from the rendered markdown/HTML instead.
• No application/ld+json structured data. REA omits schema.org RealEstateListing markup. You cannot shortcut extraction via JSON-LD parsing. Markdown scraping is the path.
• Address rendering quirk — h1 has no space between street and suburb. The h1 reads literally 6 Stawell StreetRichmond, VIC 3121 — split on the suburb token or the comma, not on whitespace.
• Apartment / unit slug prefix. 5/19 River Street becomes unit-5-19-river-st-... on the Property Pages route — not 5-19-river-st. Without the unit- prefix REA serves the wrong dwelling (the houseside #5 instead of the unit-5 inside #19).
• Off-market pages show estimated value but it's gated behind sign-in. ## Property value renders realEstimate™ $X,XXX,XXX Sign in to unlock. Don't try to log in (read-only rule); record realEstimate as gated: true and skip the numeric value.
• No history available is a valid outcome. Some addresses have never traded since REA started tracking sales. Emit historical_sales: [] rather than failing.
• Sold-listing detail pages link back to Property Pages via [Property history](/property/{slug}/) — follow that link to get the slug rather than rebuilding it; REA's own slug computation is authoritative (handles unit prefixes, street-suffix edge cases like "The Boulevard" → the-boulevard).
• Page-view counter ("429 page views") and Property ID are visible at the bottom of for-sale listings — useful as stable IDs for deduping but not required for extraction.
• Agent phone numbers are masked to the first 7 digits + ... (e.g. 0411863...) unless the user clicks "Call". Don't click. Emit the masked string as agent_phone_masked.
• "Indicative price" is mandatory in VIC under the Statement of Information regulations — when the headline says "Contact Agent" or "Auction", look for the Indicative price: $X - $Y sub-line; that's the seller's stated range and is what most agents and buyers anchor to.

Expected Output

{
  "success": true,
  "listing_url": "https://www.realestate.com.au/property-house-vic-richmond-151242968",
  "property_url": "https://www.realestate.com.au/property/6-stawell-st-richmond-vic-3121/",
  "property_id": 151242968,
  "address": "6 Stawell Street, Richmond, VIC 3121",
  "suburb": "Richmond",
  "state": "VIC",
  "postcode": "3121",
  "lat": -37.82050022,
  "lon": 145.01036193,
  "property_type": "House",
  "listing_status": "for_sale",
  "price_display": "Contact Agent",
  "indicative_price_display": "$2,200,000 - $2,400,000",
  "price_low": 2200000,
  "price_high": 2400000,
  "bedrooms": 3,
  "bathrooms": 2,
  "parking": 0,
  "land_size": "268 m²",
  "floor_area": null,
  "year_built": null,
  "agent": "Elliot Gill",
  "agent_phone_masked": "0411863...",
  "agency": "Jellis Craig - Richmond",
  "agency_url": "https://www.realestate.com.au/agency/jellis-craig-richmond-XFGDKV",
  "auction_at": "2026-06-13T11:00:00+10:00",
  "historical_sales": [
    { "year": 2018, "date": "2018-06-14", "price": 1330000, "agency": "BigginScott - Richmond" },
    { "year": 2012, "date": "2012-10-12", "price": 820000, "agency": "Belle Property - Richmond" }
  ],
  "error_reasoning": null
}

Distinct outcome shapes:

// Sold listing (post-settlement)
{
  "success": true, "listing_status": "sold", "sale_price": 815000, "sale_date": "2026-05-02",
  "price_display": "$815,000", "indicative_price_display": null, "historical_sales": [ ... ],
  "address": "5/19 River Street, Richmond, VIC 3121", "property_type": "Apartment", ...
}

// Off-market address (Property Pages, no active listing)
{
  "success": true, "listing_status": "off_market", "price_display": null, "listing_url": null,
  "realestimate_gated": true, "rental_estimate_weekly": 1285,
  "historical_sales": [],   // or populated if past sales exist
  ...
}

// No history available
{ "success": true, "historical_sales": [], "history_note": "No history available", ... }

// Slug not found / address typo
{ "success": false, "error_reasoning": "address_not_found", "did_you_mean": "6a-stawell-st-richmond-vic-3121" }

// Kasada captcha interstitial (rare with stealth on)
{ "success": false, "error_reasoning": "kasada_captcha_wall", "advice": "Re-run with --verified --proxies; rotate session." }