AU & NZ Real Estate Agent OSINT - Profiles & Decision Scores
Pricing
from $2.00 / 1,000 agent reports
AU & NZ Real Estate Agent OSINT - Profiles & Decision Scores
Build OSINT intelligence reports on real estate agents across Australia and New Zealand. Extract profiles, contacts, agency details, active listings, sales history, reviews and media mentions. Each report includes an evidence-backed 0โ100 decision score. Structured JSON for AI agents, APIs and MCP.
Pricing
from $2.00 / 1,000 agent reports
Rating
0.0
(0)
Developer
Abot API
Maintained by CommunityActor stats
1
Bookmarked
2
Total users
2
Monthly active users
5 days ago
Last modified
Categories
Features
Share
Real-Estate Agent OSINT: Scored Agent Intelligence for Australia & NZ
Real-Estate Agent OSINT turns public real-estate data into one scored, explainable report per agent. Give it an agent's name, a profile link, or a suburb, and it merges the agent's identity across sources, then combines their sold track record, contact details, vendor reviews, agency history, licence and disciplinary status and co-listing network into a single record with a transparent 0-100 score and an interactive network graph. Every value is structured, explainable data, ready for your own AI agent, app or spreadsheet to read. Built for Australia and New Zealand first.
Why This Scraper?
- One agent, every source. Portal profiles, sold history and government licence registers are merged into a single identity , with a
matchConfidenceand named evidence for every cross-site link (hard-key: a shared mobile number or email;name+office: the same name at the same office;name+suburb: the same name in a shared suburb;single-source: only one source record), so you always know what was joined and why. Two people with the same name at different offices of one franchise brand are never merged, and an office phone number shared by colleagues never links two agents. - A score you can argue with. Four pillars (performance, results, reputation, experience and trust) are each a percentile inside a peer group of the agents scored in the same run: the agent's own suburb when the run holds at least 20 agents there, otherwise one pool of the run's smaller suburbs (which can span suburbs and states).
score.peerGroupnames the group and its size. A Bayesian-shrunk rating means ten 5-star reviews can never outrank a 1,000-sale agent at 4.5 stars, and a disciplinary finding caps the score outright , it is never averaged away. - Every state, plus New Zealand. Agents anywhere in Australia (NSW, VIC, QLD, SA, WA, TAS, NT and the ACT) are found and profiled through the Domain agent directory and RateMyAgent, New Zealand agents through RateMyAgent, and Western Australian agents also through the REIWA directory. One agent found on several of these is merged into one record and billed once.
- Contacts where the source publishes them. Domain profiles carry the agent's public mobile or office number; REIWA profiles (WA) carry phone, email, office address, social links, awards and agency history. RateMyAgent contributes agency, coverage suburbs, sold history and reviews but no contact details (it does not show them openly on its own pages, so they are not collected). An email address is output only where the site shows it openly on its own pages (REIWA profiles); Domain listing rows in URL mode carry the listing agents' mobile numbers, never an email.
- Reviews without the noise. Review ratings, volumes and recommendation counts are pulled per agent, Bayesian-shrunk so that ten 5-star reviews can never outrank a proven track record, and analysed with anomaly indicators (burst patterns, near-duplicate texts) that reduce confidence instead of making accusations. Reviewer names are never published (a sign-off at the end of a review text, such as
Vendor, Jane & Johnor- Sam, is removed). Office and brand-level reputation stay clearly separate from the agent's own rating. - The network behind the agent. Co-listing data reveals who each agent really works with , an interactive
network.htmlgraph is saved to the run's key-value store whenever at least one reported agent has a co-listing partner (another agent named as lead or supporting agent on the same sale; the partner does not have to be part of the run). - Made for AI agents to read. The actor writes no prose of its own. Every number is computed deterministically and comes with its evidence (
matchEvidence,score.drivers,reviews.themesquotes,agencyGroup.matchedOn, review counts,licence.sourceUrl,profileUrls), so the AI agent that reads the dataset can quote it, check it and write its own summary.
Use Cases
- Agency principals and recruiters: find rising stars, verify licence status, and see agency-move history and key co-listing partners before a headhunt.
- Vendors choosing an agent: compare agents inside one suburb on sold volume, days on market, verified reviews and licence standing , with the evidence attached.
- Proptech and lead-gen builders: enrich an agent database with cross-site identities, sold counts and contact details through the API.
- Mortgage brokers, conveyancers and stylists: build B2B prospect lists of the top-performing agents in a territory.
- AI and MCP agent builders: a scored "who's the best agent in suburb X" endpoint with an explainable answer your agent can quote.
Data You Get
Sample shape: values are illustrative placeholders with fake data, not a real person.
| Field | Example |
|---|---|
agentId | "agent_jane_sample_a1b2c3d4" |
name | "Jane Sample" |
title | "Director / Licensee" |
phones / emails | ["61400000000"] / ["jane.sample@example.com"] |
agencyName / agencyGroup | "Ray White Sample Town" / {"group": "Ray White", "matchedOn": "Ray White", "confidence": 1} (the franchise or brand group read from the agency name, null for an independent agency) |
memberships | agency history with start/end dates where sources expose them |
officeAddress | "1 Sample Street, Applecross WA 6153" (office street address, where the source publishes it) |
socialLinks | {"facebook": "https://www.facebook.com/...", "linkedin": "https://www.linkedin.com/in/..."} (the agent's own profiles) |
suburbs / region | ["Bondi NSW"] / "Bondi NSW" |
profileUrls | links to each source profile |
soldSummary | {"count12m": 34, "count24m": 61, "volume12m": 52000000, ...} |
soldHistory | full sold listings: address, date, price where published, method |
soldHistoryComplete | true when every sold-history page was read; false when a page could not be read even after a retry, or when the agent has more sales than the per-agent limit covers (400 sales per RateMyAgent profile, 400 per Domain profile), so the list is partial: the sold analysis and the sales part of the score are left out and sold-history is not billed. The same sale published by two sources is listed once |
reviews | count, average, Bayesian rating, confidence, anomalies, recent quotes (reviewer names removed), and themes: one entry per aspect the review texts talk about (communication, negotiation, price/valuation, professionalism, honesty/trust, local knowledge, marketing/photography, responsiveness, stress/support, results/speed of sale) with count, positive / negative / neutral, positiveShare, fromRating and a sampleQuote (plus a counterQuote of the other polarity when there is one). byRole and themes cover the collected review texts only (textsAnalysed), which can be a sample of the full count |
licence | register, number, type, status, expiry, disciplinary findings + source link; attached only when the register's own address (suburb or postcode) confirms the agent, otherwise null |
licenceCandidates | unconfirmed leads: up to three register entries with the agent's name that the register's address does not confirm. Each has state, register, licenceName (as the register spells it), licenceNumber, class, status, expiry, registerLocality, matchBasis ("name-only", or "name-variant" when the entry was found through a short or full form of the given name, named in nameVariant such as "Nick -> Nicholas"), confirmed: false and contested (true when another register entry or another agent of the run shares the name or the number). Empty when the licence is confirmed |
licenceCheck | how the licence lookup ended: matched, withheld (a register match on the name alone that could be a namesake; not attached, listed in licenceCandidates when useful, matchEvidence says why), no-match, not-searched (the agent's state is unknown), no-public-register (VIC, SA, ACT, TAS, NT) or unread (the register did not answer this run) |
unreadSections | parts of the record a source did not serve this run (refused or timed out): reviews (review texts), statistics, suburbs, licence, profile (only the directory card came back). Empty when everything was read |
queryLabel | the agent name or suburb from your input that found this agent |
score | {"overall": 78, "pillars": {...}, "confidence": 0.82, "peerGroup": "bondi (34 agents)", "drivers": [...]}; each pillar lists its own drivers and a reason when it is not scored |
network | partnerCount, inRunPartnerCount and the top co-listing partners, each with coListings, lead/support split, suburbs, inRun, agentId, match and agency + agencySource (graph in network.html) |
matchConfidence / matchEvidence | 0.98 / which rule linked each source |
sources | ["reiwa", "rma", "domain", "rea"] |
changeType / changedFields / firstSeenAt / lastSeenAt | incremental-mode bookkeeping (with incremental mode off every agent row is NEW) |
The score record breaks down as pillars.performance, pillars.results, pillars.reputation and pillars.experience (each a percentile with its own coverage confidence), plus score.confidence, the peer group and the evidence that moved it (drivers).
How to read the evidence
Everything below is computed by fixed rules from the collected data, with no language model, so the same input gives the same values and every value points back to its evidence.
agencyGroup: the franchise or brand network read from the agency name against a fixed table of Australian and New Zealand networks.matchedOnis the part of the name that matched, as printed.confidenceis 1 for a distinctive brand name and lower where the brand word is also an ordinary word (for examplePlaceorActon, matched only as the first word and only in the state where that network trades). An independent agency getsnull, never a guess.reviews.themes: an aspect is listed when at least two reviews mention it (one when fewer than five texts were collected). Each mention's polarity comes from the cue words of the sentences that mention it, with negation read: "not very communicative" is negative, "never felt pressured" and "no stress" are positive, and set phrases such as "couldn't be happier" are read whole. A sentence without a cue word takes the review's star rating, andfromRatingcounts those mentions. Each quote is one real sentence of one review, at most 140 characters, cut at a word with an ellipsis, with the agent's name masked; itsdate,role,ratingandbasissay where it comes from. Sarcasm and irony are out of scope: text is read literally.score.driversandscore.pillars.*.drivers: each driver names themetric, itsvalue, itspercentileamong the agents of the same peer pool, itsweightin the pillar, itssource, andimpact, the points it moves the overall score away from a neutral peer (50th percentile, no bonus). Together with the neutral peer the impacts add up to the score.score.driversholds the five largest across the counted pillars. A pillar with no data says why inreason: the results pillar compares the agent's published days on market with at least three peers publishing the same figure, and otherwise says that no figure is published or that there is no comparison pool. List prices are not published by the sources, so there is no sale-to-list ratio.licenceCandidates: leads, never findings. Aname-variantcandidate was found only through a short or full form of the given name, or a middle given name (nameVariant, for example"Tim -> Timothy"or"Alyssa -> Ya-Yin Alyssa"), only when no register entry carries the agent's own name. Candidates never carry disciplinary data, never change the score, and onlylicenceis a confirmed licence.network: a partner is a name on the agent's own sales.inRun: truemeans it is another agent of the same run, tied by the sale record's own agent code (match: "agent-code") or by exactly one agent of that name in the sale's suburb ("name-and-suburb"). Any other partner stays a name: itsagency(agencySource: "sale-record") and rating are what the sale record shows, unconfirmed. Co-listing names come from the review site's sold records; the agent directory's sold records name no co-agents.
Coverage by region
Measured on a fixed live corpus (2026-10-22, local runs through Apify Proxy RESIDENTIAL): one suburb per Australian state and territory (two in QLD, WA and the NT) and six New Zealand suburbs (Auckland, Wellington, Christchurch, Hamilton, Dunedin, Tauranga, plus one more over the default AU exits), four agents each, 72 agent rows. Percentages are the share of reported agents that carry the value. Every row has a
scorewith at least one driver.
| Region | Sources | Identity (name, agency, suburbs) | Sold history with price and date | Reviews (count and stars) | Phone | Licence | |
|---|---|---|---|---|---|---|---|
| NSW | Domain + RateMyAgent | 100% | 100% | 100% | 100% | none | NSW register: confirmed or candidate 100% (3 of 4 confirmed) |
| VIC | Domain + RateMyAgent | 100% | 100% | 100% | 100% | none | no public register (licenceCheck: no-public-register) |
| QLD | Domain + RateMyAgent | 100% | 100% | 75% | 100% | none | QLD register, candidates only (it publishes no address): 88%, two of them found through the full form of the given name |
| SA | Domain + RateMyAgent | 100% | 100% | 75% | 75% | none | no public register |
| WA | Domain + RateMyAgent + REIWA | 100% | 100% | 88% | 100% | 63% (REIWA profiles) | WA register: confirmed 63%, below the 70% target |
| TAS | Domain + RateMyAgent | 100% | 100% | 100% | 100% | none | no public register |
| NT | Domain + RateMyAgent | 100% | 100% | 75% | 88% | none | no public register |
| ACT | Domain + RateMyAgent | 100% | 100% | 75% | 100% | none | no public register |
| NZ | RateMyAgent NZ | 100% | dated 100%, with a price 55% to 79% (New Zealand sales often withhold the price; 55% of 121 sales on a Takapuna and Sumner run, October 2026) | 100% | none (no open source publishes NZ agent phones) | none | NZ register: confirmed or candidate 86% (68% confirmed) |
| Region | Agency group (agencies of a known network) | Review themes (agents with 3+ review texts) | Score drivers | Results pillar scored | A co-listing partner named / one that is an agent of the same run |
|---|---|---|---|---|---|
| NSW | 4 of 4 | 4 of 4 | 100% | 100% | 100% / 50% |
| VIC | 4 of 4 | 4 of 4 | 100% | 100% | 100% / 50% |
| QLD | 5 of 5 | 6 of 6 | 100% | 100% | 75% / 25% |
| SA | none of the 4 agencies is a network | 2 of 2 | 100% | 75% | 25% / none |
| WA | 1 of 1 | 7 of 7 | 100% | 100% | 75% / 25% |
| TAS | 2 of 2 | 4 of 4 | 100% | 100% | 75% / 50% |
| NT | 3 of 3 | 5 of 5 | 100% | 88% | 88% / none |
| ACT | 1 of 1 | 3 of 3 | 100% | 100% | 75% / none |
| NZ | 18 of 18 | 28 of 28 | 100% | 100% | 93% / 21% |
Targets: identity 95%, sold history with price and date 85% where the market publishes prices, reviews 75%, phone 60% (Australia), score 100%, licence confirmed or candidate 70% where a public register exists, agency group on 80% of agencies that belong to a known network (no agency given a group it does not belong to: 0 of 72 rows), themes on 80% of agents with three or more review texts, drivers on every row. Below target, honestly: the WA register has no individual licence for three of the eight corpus agents: one works under the agency's body-corporate licence (the register lists the company under the agency's trading name, which is not a personal licence and is not reported as one), and two are not on the register under any form of their given name. Reviews and phones below 100% are agents who have none published on any source; a results pillar is missing where an agent publishes no days-on-market figure. Co-listing partners come from the review site's sold records (the agent directory's sold records name no co-agents), so an agent known only to the agent directory has none. Contacts follow one policy everywhere: a phone only where the Domain profile publishes it, an email only where a REIWA profile shows it openly, and never the contact details RateMyAgent hides on its own pages. A person a suburb sweep finds on only one of the two Australian directories is looked up on the other and merged only when it is the same name at the same office (or the same office spelled differently) and the merge's identity rule joins the two records. Each four-agent suburb took about 90 to 250 seconds.
How to Use
- Pick a mode. Realestate (default) profiles agents from agent queries (name + suburb), profile links or a suburb sweep. URL scrapes the links you paste into
urls(see "URL mode shapes" below). When URL mode is selected,urlsis used and agent queries, profile links and suburbs are ignored; in Realestate modeurlsis ignored. - Switch report sections on or off: sold history, reviews, licence check, network.
- Set Max results per run to control cost, then click Start.
The run first collects every agent, then scores them all together and delivers the rows. Rows arrive in the dataset after collection finishes, not one by one as agents are found; while it collects, the run's status message shows how many agents were found so far. Every agent is scored against the full set of agents collected in the run, so the order in which agents are found never changes a score or a rank; the peer group size is shown in score.peerGroup.
Profile agents by name:
{"agentQueries": [{ "name": "Jane Sample", "suburb": "Bondi NSW" }],"maxItems": 5}
Sweep a WA suburb (the prefilled Console example: 4 agents, about two and a half to three minutes on the Apify platform):
{"suburbs": ["Applecross WA"],"maxAgentsPerSuburb": 4}
Sweep suburbs in other states and territories (Domain and RateMyAgent directories, about 25 to 30 seconds per agent on the Apify platform, so 12 agents take about five to six minutes; a city name resolves to its centre, so "Darwin" is Darwin City NT 0800):
{"suburbs": ["Bondi NSW", "Sandy Bay TAS", "Darwin"],"maxAgentsPerSuburb": 4}
Sweep New Zealand suburbs (RateMyAgent NZ and the NZ licence register):
{"country": "nz","suburbs": ["Ponsonby, Auckland", "Thorndon, Wellington", "Riccarton, Canterbury"],"maxAgentsPerSuburb": 4}
Add "suburbLeaderboard": true to also rank each suburb's agents (a board needs at least three scored agents). For New Zealand set "country": "nz" (RateMyAgent NZ and the NZ licence register). New Zealand runs work over the default AU residential exits as well as over NZ exits ("apifyProxyCountry": "NZ" in proxyConfiguration); both were measured on the same suburbs with the same rows returned. To profile specific agents, paste their profile links into agentUrls (domain.com.au /real-estate-agent/<name>-<id>, ratemyagent /real-estate-agent/<name>-<code>/..., reiwa.com.au or realestate.com.au).
URL mode: set "mode": "url" and keep the prefilled urls, or replace them with your own links (for example "urls": ["https://reiwa.com.au/for-sale/applecross/"] with "maxItems": 20); see "URL mode shapes" below.
Output views
The dataset opens on Overview (every row of every kind, the view the CSV and Excel export uses), then a set of column layouts: Agents (the full scored report columns, view default), Listings (property columns of URL mode), Agencies (agency columns of URL mode), Articles (news columns), Leaderboard (suburb ranking columns), Changes (incremental changeType and changedFields) and Sold history (view sales: one row per sale, with the agent's id, name and agency). Views pick columns, they do not filter rows: every view lists every row of the dataset, and a row of another kind shows those columns empty (recordType and changeType tell the kinds apart; in the sales view a row without a sold history stays one row with the sale columns empty). Every field of every row kind is declared with its type and a description in the dataset schema, and the platform checks each row against it when it is saved. A run summary is saved to the key-value store as OUTPUT on every run, a failed or empty one included: {"schemaVersion": "1.0", "mode": "suburbs", "agentsReported": 4, "partial": false, "sourcesNotRead": 0, ...} with rows reported, identities merged, refused sources, the co-listing edges (networkEdges) and suburbResolution (for each suburb you typed: status resolved / ambiguous / not-found / refused, the suburb it was resolved to, the candidate suburbs and a countryHint). schemaVersion is the version of the output contract: a minor step only adds fields, a major step renames, removes or retypes one, so an AI agent or a pipeline can check it before reading rows. Rows carry no version field of their own.
Run it from your code
Python:
from apify_client import ApifyClientclient = ApifyClient("<YOUR_APIFY_TOKEN>")run = client.actor("abotapi/realestate-agent-osint").call(run_input={"agentQueries": [{"name": "Jane Sample", "suburb": "Bondi NSW"}],})for agent in client.dataset(run["defaultDatasetId"]).iterate_items():print(agent["name"], agent["score"]["overall"], agent["phones"])
JavaScript:
import { ApifyClient } from 'apify-client';const client = new ApifyClient({ token: '<YOUR_APIFY_TOKEN>' });const run = await client.actor('abotapi/realestate-agent-osint').call({agentQueries: [{ name: 'Jane Sample', suburb: 'Bondi NSW' }],});const { items } = await client.dataset(run.defaultDatasetId).listItems();
Or connect it to Make, Zapier, n8n, Google Sheets or webhooks from the Integrations tab.
Use with AI agents
The actor is built to be called by an AI agent, as a tool through Apify's MCP server or through the API, and to return data the agent can reason over and cite. It writes no prose of its own. The suburb sweep and the New Zealand MCP call below were run verbatim on the Apify platform (October 2026). The name lookup uses the placeholder name Jane Citizen: put the name of the agent you want to look up; its output was produced by the same call with a real agent's name. Every output below is an illustrative, trimmed sample of a real row, with the agent's identity and contact fields masked.
Python, look up one agent in a suburb:
from apify_client import ApifyClientclient = ApifyClient("<YOUR_APIFY_TOKEN>")run = client.actor("abotapi/realestate-agent-osint").call(run_input={"agentQueries": [{"name": "Jane Citizen", "suburb": "Paddington QLD"}],"maxItems": 1,})for agent in client.dataset(run["defaultDatasetId"]).iterate_items():print(agent["name"], agent["agencyName"], agent["score"]["overall"], agent["licenceCheck"])
Trimmed output of the name lookup (illustrative; identity masked):
// stdout: [agent name] Ray White Paddington QLD 49 withheld// dataset row (cloud run, October 2026), trimmed; identity masked{"agentId": "[masked]","name": "[masked]","agencyName": "Ray White Paddington QLD","agencyGroup": { "group": "Ray White", "matchedOn": "Ray White", "confidence": 1 },"region": "Paddington QLD","sources": ["rma", "domain"],"matchConfidence": 0.96,"matchEvidence": [{ "rule": "name+office", "detail": "rma+domain: name similarity 1.00 at the same office (Ray White Paddington QLD)", "confidence": 0.96 },{ "rule": "licence-withheld", "detail": "unconfirmed name-only match withheld: the QLD register lists licensee(s) of this name, but its own locality does not confirm any of the agent's suburbs", "confidence": 0 }],"licence": null,"licenceCheck": "withheld","licenceCandidates": [{ "state": "QLD", "register": "QLD Office of Fair Trading", "licenceName": "[masked]", "licenceNumber": "[masked]", "class": "REAL ESTATE AGENT","status": "Current", "registerLocality": null, "matchBasis": "name-only", "confirmed": false, "contested": true }],"soldSummary": { "countAll": 41, "count12m": 41, "volume12m": 72279000 },"soldHistoryComplete": true,"reviews": { "count": 182, "average": 4.9, "bayesian": 4.91, "textsAnalysed": 30 },"unreadSections": [],"score": { "overall": 49, "confidence": 0.81, "peerGroup": "region pool (1 agent)","drivers": [{ "pillar": "reputation", "metric": "review-recency", "label": "latest review 2024-06-10 (recency 0.21 on a 3-year scale)", "impact": -1.53 }] }}// key-value store record OUTPUT{ "schemaVersion": "1.0", "mode": "queries", "agentsReported": 1, "partial": false, "sourcesNotRead": 0, "networkEdges": { "named": 7, "inRun": 0 } }
JavaScript, sweep a suburb and rank its agents:
import { ApifyClient } from 'apify-client';const client = new ApifyClient({ token: process.env.APIFY_TOKEN });const run = await client.actor('abotapi/realestate-agent-osint').call({suburbs: ['Paddington QLD'],maxAgentsPerSuburb: 4,suburbLeaderboard: true,});const { items } = await client.dataset(run.defaultDatasetId).listItems();for (const row of items) {if (row.changeType === 'LEADERBOARD') console.log('Leaderboard', row.region, row.score.map((r) => `${r.rank}. ${r.overall}`).join(' | '));else console.log(row.agencyName, row.score?.overall, row.licenceCheck, row.soldHistoryComplete);}
Trimmed output of the suburb sweep (illustrative; agent names masked):
// stdout (agency, score, licenceCheck, soldHistoryComplete per agent, then the board)// Ray White Paddington QLD 42 withheld true// Wakely Properties 47 withheld true// Place West 70 withheld true// Ray White Paddington QLD 44 withheld true// Leaderboard Paddington QLD 1. 70 | 2. 47 | 3. 44 | 4. 42// the leaderboard row (cloud run, October 2026), agent names masked{"agentId": "leaderboard_paddington","changeType": "LEADERBOARD","region": "Paddington QLD","country": "au","score": [{ "rank": 1, "agentId": "[masked]", "name": "[masked]", "agency": "Place West", "overall": 70, "scoreConfidence": 0.67 },{ "rank": 2, "agentId": "[masked]", "name": "[masked]", "agency": "Wakely Properties", "overall": 47, "scoreConfidence": 0.71 }],"sources": ["rma", "domain"]}// the top driver of the first-ranked agent's row{ "pillar": "performance", "metric": "sales-count", "label": "49 sales in the last 12 months: 88th percentile in the region pool of 4" }
MCP tool call for Apify's MCP server (mcp.apify.com), a New Zealand suburb:
{"jsonrpc": "2.0","id": 1,"method": "tools/call","params": {"name": "call-actor","arguments": {"actor": "abotapi/realestate-agent-osint","input": { "country": "nz", "suburbs": ["Ponsonby, Auckland"], "maxAgentsPerSuburb": 3 },"waitSecs": 45}}}
The call returns the run (this one took 72 seconds); the agent then calls get-actor-run until it has finished and reads the rows with get-dataset-items (datasetId from the run, for example fields: "name,agencyName,score,licenceCheck,licenceCandidates,matchEvidence").
Trimmed output of the New Zealand call (illustrative; identity masked):
// get-dataset-items, first row (cloud run, October 2026), trimmed; identity masked{"name": "[masked]","agencyName": "Bayleys Real Estate Ponsonby","licenceCheck": "matched","licence": { "register": "Real Estate Authority (NZ)", "jurisdiction": "NZ", "licenceNumber": "[masked]", "status": "Active","disciplinary": { "flagged": false, "items": [] } },"licenceCandidates": [],"matchEvidence": [{ "rule": "single-source", "detail": "one rma record; no other source matched this agent", "confidence": 0.6 }],"score": { "overall": 66, "peerGroup": "region pool (3 agents)","drivers": [{ "pillar": "results", "metric": "days-on-market", "label": "average 21 days on market (review site): 83rd percentile in the region pool of 3, lower is better", "impact": 5.69 }] }}
How an agent should read the evidence:
matchEvidence: why the sources were merged into one person.hard-key(a shared mobile number or email) is the strongest;name+officeandname+suburbcarry the name similarity;single-sourcemeans only one site knew the agent.matchConfidenceis the strongest rule. Quote thedetailsentence when you say two profiles are the same person.licenceCheckandlicenceCandidates:licenceis set only whenlicenceCheckismatched. AlicenceCandidatesentry is a lead, not a fact: a register entry with the same name that the register's own address does not confirm (confirmed: false;contested: truewhen another entry or another agent shares the name or number). Never present a candidate as the agent's licence, and never read disciplinary data into it (it carries none).no-public-registermeans the state has no searchable register (VIC, SA, ACT, TAS, NT),not-searchedthat the agent's state is unknown, andunreadthat the register did not answer this run, which says nothing about the licence.score.drivers: the evidence behind the number. Each driver has alabelsentence ready to quote, itsvalue,percentileandsource, andimpact, the points it moves the score from a neutral peer. Scores are percentiles insidescore.peerGroup, the agents of the same run, so compare scores only between rows of one run.unreadSections: parts a site did not serve on this run (for examplereviews). A value missing because its part is listed here is unknown, not zero or empty; run again later before concluding anything from it.soldHistoryComplete:truemeans the sold list is every sale the sources publish (about the last 12 months).falsemeans the list is partial:soldAnalysisisnull, so do not compute totals from it.poolPartialandOUTPUT.partial: when the run's time limit cut collection short,OUTPUT.partialistrueand every score, network partner and percentile is relative to that smaller pool (in incremental mode such a stored record is markedpoolPartialand refreshed by the next complete run without a charge). Prefer a complete run before ranking agents.- A value no source publishes stays
nullor an empty list. Readnullas "not published", never as "none".
Resume and recurring updates
- Resume (
resumeFromRunId) continues one interrupted run: paste its run ID (or its dataset ID) and the actor skips agents already reported there, so you don't pay twice. - Incremental mode (
incrementalMode) is for scheduled runs over the same scope. With it off (the default) the run reads and writes no stored state and every agent row isNEW. With it on, each record is classifiedNEW,UPDATED(withchangedFields),UNCHANGED(suppressed and not billed unlessemitUnchangedis on) orEXPIRED(only withemitExpired).stateKeynames or shares the stored state. Compared: name, contact details, agency, licence, sold-listing summary and history, and review statistics (count and average). Excluded by default but still present in the output:fetchedAt,lastSeenAt, photo URL, recent review text and the review-text sample size (reviews.textsAnalysed), source lists, the run bookkeeping fieldsqueryLabel,unreadSectionsandlicenceCheck, the unconfirmedlicenceCandidates(theircontestedflag depends on which other agents the run found), and the three values that are relative to the other agents of the run:score,networkand the Bayesian ratingreviews.bayesian. Those three move whenever another agent joins, drops out or could not be read, so they never make an agentUPDATEDon their own (and never cause a charge); a real change of the agent, such as a new sale, a new review or a licence change, still does. The stored snapshot is re-compared with the current ignore set on every run, so adding an ignored field never causes a wave ofUPDATEDrows.EXPIREDis emitted only after a complete suburbs-only scan: never after amaxItemscap hit, amaxAgentsPerSuburbtrim, a run cut short by its time limit, pasted URLs or name queries. A sold list a site did not serve in full (a page refused or timed out) is never compared against, nor stored over, the agent's previous one, so it does not turn an unchanged agent into anUPDATEDrow; a list longer than the per-agent limit is read in full every run and compared as it is (a new sale isUPDATED). Likewise, when a source that listed an agent on the stored run REFUSES on a later run (or does not answer in time), the stored record is kept as it is: that agent is not reported or billed on that run, it is not markedEXPIRED, and the next run with every source back compares against the untouched stored record. A part of a record a source did not serve (review texts, statistics, the licence register; seeunreadSections) is kept from the stored record the same way. This waiting is bounded: after two runs in a row, the record as served is taken as the agent's record (a source that stopped answering for good cannot freeze an agent). A source that answers and no longer lists the agent is a real change and is reported. Any refused profile, suburb list or name search makes the scan incomplete, so no agent is markedEXPIREDon that run, and the run's status and OUTPUT (sourcesNotRead) say how many could not be read. A record first stored from a partial read (for example a sold list cut short) is refreshed by the next complete read without anUPDATEDrow or a charge.lastSeenAtof an unchanged agent is refreshed in the stored state on every run that sees it.
Send results into your apps (MCP connectors)
Optionally pipe the results into the apps you already use, via Model Context Protocol (MCP) connectors. This is an extra delivery step after the run: the Apify dataset is never changed.
What gets written to the connector: a condensed, human-readable summary of each record, not the full JSON. Each item becomes one entry with a title and its key fields flattened to plain text. The complete record always stays in the Apify dataset.
- Authorize a connector once under Apify โ Settings โ Integrations (Notion, Linear, Airtable, or Apify).
- Select it in the "Pipe results into your apps" input field. (If the picker is empty, you haven't authorized a connector yet.)
- For Notion, also set
notionParentPageUrlto the page where items should be created.
The connection is mediated by Apify's MCP proxy, so this actor never sees your third-party credentials. Leave the field empty to skip.
Input Parameters
| Parameter | Type | Default | Description |
|---|---|---|---|
agentQueries | array | [] | {"name": "...", "suburb": "..."} pairs to profile. Searched on the Domain agent directory (Australia) and RateMyAgent (Australia and New Zealand); a WA suburb (or no suburb) also searches the REIWA directory. The suburb (with its state, e.g. "Bondi NSW") keeps namesakes out: it is resolved like a sweep's suburb (see How is a suburb name matched?), the directories are searched in that state, and a match whose profile does not cover that suburb is skipped. An ambiguous or unknown suburb is not searched and not billed, and the run names the candidates. |
agentUrls | array | [] | domain.com.au (/real-estate-agent/<name>-<id>), RateMyAgent (/real-estate-agent/<name>-<code>/...), reiwa.com.au or realestate.com.au agent profile links. |
suburbs | array | [] | Suburbs to sweep for all agents. Australia: best with state and postcode ("Brighton VIC 3186", "Bondi NSW", "Sandy Bay TAS"); New Zealand (country nz): with the region ("Ponsonby, Auckland", "Thorndon, Wellington", "Richmond, Tasman"). Each suburb is resolved once and every directory sweeps that same suburb (see How is a suburb name matched? below). Swept on the Domain directory (Australia) and RateMyAgent (Australia and New Zealand); Western Australian suburbs ("Applecross WA") also on the REIWA directory. A person listed by only one of the two directories is looked up on the other by name and added only when it is the same person at the same office. |
maxAgentsPerSuburb | integer | 25 | Cap per suburb during a sweep, counted as people across the directories (the same agent listed by two directories is one agent and one record). |
mode | enum | realestate | realestate (agent OSINT) or url (scrape pasted links). |
urls | array | 16 example links | URL mode only. Prefilled with one example of every confirmed listing, agency, directory and article shape (no individual agent's profile); replace them to scope a run. maxItems is split evenly across the links. |
country | enum | au | au (Australia: NSW, VIC, QLD, SA, WA, TAS, NT, ACT) or nz (New Zealand, every region). A suburb that only the other country has is not swept; the run says which country to set. |
includeSoldHistory | boolean | true | Attach the full sold track record per agent. |
includeReviews | boolean | true | Fetch and analyse vendor reviews. |
includeLicenceCheck | boolean | true | Check government licence registers: NSW, QLD and WA in Australia and the national register in New Zealand. Only the register of the agent's own state is searched (the state its profile suburbs name; an agent whose state is unknown or whose suburbs span several states gets no lookup). A licence is attached only when exactly one licensee matches the name AND the register's own address confirms the agent: the register's suburb is the agent's base suburb (the suburb the sweep found the agent in, else the office's suburb, else the suburb of most of their sales), the office's suburb or the swept suburb, or its postcode is the office's postcode. A suburb the agent only sold in, or lists among the markets they serve, does not confirm: such an entry is listed as a candidate. A match on the name alone is never attached (licence stays null, licenceCheck is withheld): a missing licence is better than another person's. The QLD register publishes no address, so QLD agents get withheld rather than a guess. Such name-only entries are listed in licenceCandidates instead (see Confirmed licence or candidate? below); when no entry carries the agent's own name, the short and full forms of the given name (Nick and Nicholas, Tim and Timothy) are searched too, and an entry whose middle given name is the agent's name (Ya-Yin Alyssa Liao for Alyssa Liao) is listed too; what they find is only ever a candidate. On the New Zealand register the Preferred Name column counts as the licensee's name (Rosemary Arnott, preferred name Rose, is Rose Arnott), and a name the register search cannot find is looked up by surname once. The NSW register lists at most 200 licensees per surname and the NZ register shows one page of 25, so a name that may be past the cut-off gets no licence; a licence number that matched two agents of one run is attached to neither; and when the run finds another agent of the same name in the same state whom the register's address does not rule out, the licence is attached to neither. An agent is flagged disciplined only when the register itself records an action for that licence, read only from that licensee's own register record. Victoria, South Australia, the ACT, Tasmania and the Northern Territory are not searched (no register that permits automated lookup), so their agents carry no licence record. Register lookups add a few seconds per agent. |
includeNetwork | boolean | true | Build the co-listing network (partners named on the same sales; another agent of the run is marked inRun); network.html is saved when at least one reported agent has a co-listing partner on its sold listings. |
aiAnalysis | boolean | false | Deprecated: no longer used. Kept so saved tasks still start; the value is ignored and changes nothing in the output or the price. |
suburbLeaderboard | boolean | false | Push a ranked leaderboard record per suburb. |
proxyConfiguration | object | Residential, AU | Connection used for the agent directories and profile pages. The Domain and RateMyAgent sources need a residential connection. New Zealand runs work over the default AU exits too (measured), or set apifyProxyCountry to NZ. With the group set to Automatic, those sources still use the RESIDENTIAL group and the run says so in its log. See Plan Requirement. |
maxRequestsPerMin | integer | 24 | Polite pacing cap, per site (each directory, review site and licence register separately). |
maxItems | integer | 50 | Realestate mode: stop after this many agents, counted as people (one agent found on several sources is one agent and one report; a person's records are never split by the cap), so a run can return fewer rows than the source records it read. Once the cap is reached, a further listing is fetched only when it names the office of an agent already counted. URL mode: split evenly across the links so every link is read (each link returns at most its share, at least 1 record). |
resumeFromRunId | string | (none) | Continue one interrupted run. |
incrementalMode | boolean | false | Return only new and changed agents on scheduled runs. |
stateKey | string | default | Name or share an incremental-mode state. |
emitUnchanged | boolean | false | Also return (and bill) unchanged agents. |
emitExpired | boolean | false | Also return agents gone from the source set. |
ignoreFieldsForChanges | array | [] | Ignore fields when detecting changes (incremental mode only). Extra fields, dotted paths allowed for nested ones, that must not alone mark a record UPDATED. |
mcpConnectors | array | (none) | Connectors to pipe results into (optional). |
notionParentPageUrl | string | (none) | Notion parent page (Notion connector only). |
maxNotifyListings | integer | 50 | Max items to export per connector. |
URL mode shapes
Default disclosure: in URL mode the urls field is prefilled with the 16 example links below (every row of the table except the agent-profile pattern), one per confirmed shape. If you switch mode to url and leave urls untouched, the actor scrapes all of them, not nothing. maxItems is split evenly across the links so every link is read: with the default 50 and 16 links, the first two links return up to 4 records each and the others up to 3, and a link always gets at least 1 even when you have more links than maxItems. Replace the list with your own links to scope a run; an empty list stops the run with an error.
Not supported: https://reiwa.com.au/agent-finder/ is a search form that lists no agents, so it is skipped with a SKIP warning; paste a /real-estate-agents/<suburb>/ page instead.
Other accepted links: a domain.com.au agent profile (/real-estate-agent/<name>-<id>) and a ratemyagent agent profile (/real-estate-agent/<name>-<code>/) each give one agent row, read through the same agent directories as Realestate mode (residential connection required). A realestate.com.au agent profile, a domain.com.au /listing/<id> link, and realcommercial.com.au and property.com.au search links sit behind bot protection that a plain datacenter or residential connection does not clear, so they are tried on the actor's escalation lanes and may be skipped. Listing rows carry the listing agents' names, mobile numbers and profile links, never their email addresses (see Contacts where the source publishes them above).
Unconfirmed shapes: a link the actor does not recognise or cannot read is never a silent empty result. Each one logs a visible warning of the form SKIP <url> - <reason> (for example unknown domain, no RMA agent slug/code in URL, no records extracted (page shape not supported or content empty)) and the run continues with the next link. If no link loaded any page at all, the run fails with a clear connection message instead of finishing empty.
Billing in URL mode: only agent records that carry an agency, phone, sold or review value (a reiwa agent row, or an agent profile from ratemyagent, realestate.com.au or domain.com.au) are billed, one agent-report event each. Property, article, news and listing-style rows are pushed without any charge. An agent reached by several of your links (a directory page and the agent's own profile, or the same link twice with tracking parameters) is reported and billed once per run. Other events (sold-history, reviews-multisite, network, suburb-leaderboard) apply to Realestate mode only.
Output Example
Sample shape: values are illustrative placeholders with fake data, not a real person.
{"agentId": "agent_jane_sample_a1b2c3d4","changeType": "NEW","name": "Jane Sample","title": "Director / Licensee","phones": ["61400000000"],"emails": ["jane.sample@example.com"],"agencyName": "Ray White Sample Town","agencyGroup": { "group": "Ray White", "matchedOn": "Ray White", "confidence": 1 },"suburbs": ["Bondi NSW"],"profileUrls": ["https://www.example.com/agent/jane-sample"],"soldSummary": { "countAll": 214, "count12m": 34, "count24m": 61, "volume12m": 52000000 },"soldHistory": [{ "address": "12 Sample Street, Bondi NSW", "soldDate": "2025-11-14", "price": 2450000, "method": "Private treaty", "source": "rma" }],"reviews": { "count": 96, "average": 4.6, "bayesian": 4.51, "confidence": 0.76, "textsAnalysed": 20, "anomalies": [],"themes": [{ "aspect": "communication", "count": 12, "positive": 11, "negative": 1, "neutral": 0, "positiveShare": 0.92, "fromRating": 2,"sampleQuote": { "text": "[agent] kept us informed every step of the way.", "date": "2026-08-02T10:00:00Z", "role": "VENDOR", "rating": 5, "source": "rma", "basis": "cue-words" },"counterQuote": { "text": "No follow up after the first open home.", "date": "2026-03-01T09:00:00Z", "role": "BUYER", "rating": 1, "source": "rma", "basis": "cue-words" } }] },"licence": { "register": "NSW Fair Trading", "licenceNumber": "00000000", "status": "Current", "disciplinary": { "flagged": false, "items": [] } },"licenceCheck": "matched","licenceCandidates": [],"score": {"overall": 78,"pillars": {"performance": { "percentile": 88, "confidence": 1, "reason": null, "drivers": [{ "pillar": "performance", "metric": "sales-count", "value": 34, "percentile": 91, "weight": 0.33, "impact": 4.1, "label": "34 sales in the last 12 months: 91st percentile in the bondi pool of 34", "source": "sold history (domain, rma)" }] },"results": { "percentile": null, "confidence": 0, "reason": "no days-on-market figure is published for this agent, and the sources publish no list prices (no sale-to-list ratio): nothing to compare", "drivers": [] },"reputation": { "percentile": 71, "confidence": 0.76, "reason": null, "drivers": [] },"experience": { "percentile": 64, "confidence": 0.6, "reason": null, "drivers": [] }},"confidence": 0.82,"peerGroup": "bondi (34 agents)","drivers": [{ "pillar": "performance", "metric": "sales-count", "value": 34, "percentile": 91, "weight": 0.33, "impact": 4.1, "label": "34 sales in the last 12 months: 91st percentile in the bondi pool of 34", "source": "sold history (domain, rma)" }]},"network": { "partnerCount": 12, "inRunPartnerCount": 1, "topPartners": [{ "name": "Sample Partner", "agency": "Ray White Sample Town", "agencySource": "sale-record", "inRun": false, "agentId": null, "match": null, "coListings": 9, "asLead": 2, "asSupport": 7 }] },"matchConfidence": 0.98,"sources": ["rma", "domain"],"country": "au"}
Plan Requirement
The connection requirement depends on the source:
- reiwa.com.au, Guardian and Urban Developer pages (URL mode), and the REIWA agent lane of Realestate mode (WA agents): open pages that load on a datacenter connection, no residential access needed.
- Agents outside Western Australia (and New Zealand): the Domain agent directory and RateMyAgent answer residential connections only (the default: Apify Proxy, group RESIDENTIAL; with the group set to Automatic these two sources still use RESIDENTIAL, and the run logs that once). On a run without residential access (a free plan, or datacenter-only proxies) both sources are skipped after the first refused connection, nothing from them is reported or billed, and a run that found nothing else fails with a message starting Residential connection required instead of finishing empty. A run whose listed or found agent profiles were all refused fails too (nothing reported or billed), so a refusal never looks like an empty suburb. When RateMyAgent refuses a residential connection for a page, the actor retries it on a fresh connection, then on a refreshed session, and finally (read requests only) through Apify's unblocking proxy group; that proxy traffic counts toward your Apify proxy usage like any other proxy request. When you choose a datacenter-only proxy group yourself, the unblocking group is never used (it is a different, separately billed group you did not pick): RateMyAgent is then simply skipped.
- realestate.com.au, realcommercial.com.au and property.com.au pages (profile links and URL mode): behind bot protection that plain requests do not clear; they are tried on the actor's escalation lanes and may be skipped. Government licence registers are reachable without a proxy.
Measured on the Apify platform with the prefilled Console example (Applecross WA): with a datacenter-only proxy group the run still succeeds with the 4 WA agents from the REIWA directory and uses no residential traffic and no unblocking-group traffic; it logs once that the selected group cannot reach the other directories and its final status counts the profiles and lists that could not be read (nothing from them is billed). With the proxy switched off (useApifyProxy: false) it logs one warning and returns the agents the open sources serve, with the same final count of what could not be read. Measured the same way for agents outside Western Australia (a Bondi NSW sweep of 2 agents, October 2026): with a datacenter-only proxy group the run fails after about a minute with the message starting Residential connection required, nothing reported or billed; with the proxy switched off it returned the 2 agents from the review site's directory only (no agent-directory data, so no phone numbers) and its final status counted 3 profiles, name searches or suburb lists that could not be read.
Your Apify plan must include proxy access. On a free plan the actor logs this warning once and continues on a backup connection where possible:
The selected connection is unavailable to this run. Apify free plans do not include residential proxies. Continuing on a backup connection; some pages may still be unavailable.
If no page could be read the run fails with one of these messages instead of returning an empty dataset:
- The proxy connection was refused for this account, so no page could be read. Your Apify plan does not include proxy access; a paid plan is required.
- No results: residential proxies are unavailable on this run (Apify free plans do not include them), and the backup connection could not load the requested real-estate pages.
- Could not load any results page (the requests failed), so nothing was collected. This is a connection problem, not an empty search. Please try again later.
FAQ
How much does it cost?
Pay per event: the Pricing tab shows the current rate of each event. The actor charges these events, each only after the row it belongs to is saved:
| Event | Unit | When it is charged |
|---|---|---|
| agent-report | one agent row | Once per agent row delivered in Realestate mode, and once per agent row in URL mode (a row with an agency, phone, sold or review value). Property, agency and news rows are never charged. |
| sold-history | one agent | Once per delivered agent whose sold list is complete (soldHistoryComplete: true) and holds at least 5 sales, with sold history switched on. A partial list is delivered, flagged, and not charged. |
| reviews-multisite | one agent | Once per delivered agent with at least one analysed review, with reviews switched on. Not charged when the review texts were not served on that run (unreadSections lists reviews or profile). |
| network | one run | Once per run, only when agent rows were delivered, the collection was complete (not cut short by the run's time limit) and the co-listing graph has at least one link (a partner named on a reported agent's sales), so network.html was saved. |
| suburb-leaderboard | one leaderboard row | Once per leaderboard row delivered: a suburb with at least three scored agents, with suburbLeaderboard on and a complete collection (a ranking of part of a suburb would be wrong). |
| apify-actor-start | one run start | Charged by the Apify platform when the run starts: one event per GB of run memory, at least one (a run at the default 512 MB is one event). |
Nothing else is charged. In incremental mode, UNCHANGED rows are suppressed and not charged (with emitUnchanged on they are delivered and charged like any row); with incremental mode off no stored state is read or written, and every agent row is delivered as NEW and charged. An agent is only reported and charged when at least one source returned real data for them (agency, phone, sold history or reviews). A licence register match alone does not count, because registers are searched by name and cannot tell namesakes apart. A name-only match whose profile sources all failed or came back empty is skipped with a warning in the log, never charged and never counted toward Max results per run, in both modes (in URL mode a skipped match does not use up its link's share either: the share is filled from that link's next records). A name match whose profile lists no suburb is skipped too, because it cannot be checked against the suburb you asked for. A profile, suburb list or name search a site refused is never charged (nothing of it is delivered), and in URL mode a listing or article reached from two of your links is delivered and counted once. Max results per run caps the spend of any run: at most that many agent-report, sold-history and reviews-multisite events, one network event, and one suburb-leaderboard event per suburb with three or more scored agents, on top of the run's apify-actor-start event. The run's maximum cost per run (the maximum charge you set for the run on Apify) is respected the same way: the actor stops collecting new agents once that budget cannot pay one more agent report, never delivers a row whose charges do not all fit under it (no row is ever delivered unbilled), and ends with a status that starts with Maximum cost per run reached: and says how many agents were delivered and how many were found but not delivered (OUTPUT.agentsNotDeliveredForCost). A budget that cannot pay a single agent report ends the run as failed, with nothing delivered or billed beyond the run start.
Proxy traffic is a separate cost on your Apify account, not one of these events. In a default run every agent is read over residential proxies, Western Australian agents included (their Domain and RateMyAgent profiles are residential-only; only their REIWA profile and the licence registers use none): measured 0.32 to 0.73 MB of residential traffic per agent (0.32 MB in New Zealand, 0.54 to 0.73 MB in Australia, October 2026; see How long does a run take below). A run on a datacenter-only proxy group uses no residential traffic and reads WA agents from REIWA only. Pages a review site refuses may be retried through Apify's unblocking proxy group, which counts toward your proxy usage too.
Confirmed licence or candidate?
licence is the agent's licence only when it is confirmed: a register entry carries the agent's name, the register's own address matches where the agent is based (the agent's base suburb, office suburb or postcode, or the suburb whose sweep found them; never merely a suburb they sold in), and no other agent of the same name in the run could be that licensee. Only a confirmed licence can carry disciplinary findings, and only a confirmed licence counts in the score.
When the register knows the name but cannot confirm the person (the QLD register publishes no address; in WA, NSW and NZ the registered suburb may be another one), licence stays null and up to three matching entries are listed in licenceCandidates. A candidate is a lead to check, not a finding: it is marked matchBasis: "name-only" (or "name-variant", with nameVariant such as "Nick -> Nicholas", when only a short or full form of the given name found it, because the QLD and WA registers file full given names) and confirmed: false, it never carries disciplinary or compliance details (a namesake's record must never read as the agent's), and it has no effect on the score. contested: true means another register entry or another agent in the same run shares the name or the licence number, so treat it with extra care. No candidates are listed when more than five register entries share the name, when the register list is cut off, or when the register did not answer; matchEvidence then says why.
Is it legal to scrape real-estate agent directories and licence registers?
- Public sources only. The actor reads what the agent directories, the review site and the government licence registers publish openly to every visitor. It uses no login, no account and no credentials (neither its own nor yours) and reads nothing behind a paywall.
- Licence and disciplinary data come only from the one matched register entry. A disciplinary finding is reported only when the register itself records it for the licence attached to the agent (
licence), never from a candidate:licenceCandidatescarry no disciplinary or compliance data. Every licence carries the register link to check it. - Contact data policy. A phone number or email address is output only where the source shows it openly on its own pages: the public phone of a Domain agent profile, and the phone and email of a REIWA profile. Contact details RateMyAgent masks on its own pages are never collected, and the New Zealand register's contact details are never revealed. Reviewer names are removed from every review.
- Privacy law. Agents' names, work phones and emails, sales and licences are published by or about them in their professional capacity, but they can still be personal information under the Australian Privacy Act 1988 (Australian Privacy Principles) and the New Zealand Privacy Act 2020 (Information Privacy Principles). If you store, combine or use this data to contact agents, doing so lawfully is your responsibility, including the Spam Act 2003 and the Do Not Call Register in Australia and the Unsolicited Electronic Messages Act 2007 in New Zealand. This is general information, not legal advice; get your own advice before any commercial redistribution.
- Corrections and removal. An agent who wants a record corrected or removed from this actor's output can ask through the Support section at the end of this page.
Can I get only new or changed agents on a schedule?
Yes. Schedule the actor from the Schedules tab and turn on Incremental mode with a stable stateKey. Each run then returns only new and updated agent records, and unchanged ones are not billed. Agency moves, licence changes, new sales and new reviews surface as UPDATED with changedFields. A score or network shift on its own does not, because both are measured against the other agents of the run (a peer joining or dropping out would otherwise mark every agent as changed); every delivered row still carries the current score.
Why is a suburb's leaderboard smaller than the suburb's agent count?
A suburb's board ranks the agents of this run that the sweep of that suburb found or whose main suburb it is, once they have a score, and it needs at least three of them. It ranks the agents the run collected (up to Max agents per suburb), not every agent of the suburb, and records missing a score are never ranked silently. Run a bigger sweep for a fuller ranking: scores compare agents against the other agents of the run.
How long does a run take, and how much proxy traffic does it use?
Measured on the Apify platform with the default settings (proxy group RESIDENTIAL, AU exits; 24 requests per minute per site):
| Run | Agents | Time | Memory used (peak) | Residential proxy traffic |
|---|---|---|---|---|
| Console example (Applecross WA) | 4 | 167 s | 104 MB | 1.7 MB |
| One suburb in each of NSW, VIC, QLD, WA and SA | 5 | 178 s | 102 MB | 1.9 MB |
| New Zealand (Riccarton, Ponsonby) | 8 | 183 s | 95 MB | 3.3 MB |
| QLD, TAS, NT, ACT and NSW suburbs | 20 | 544 s | 123 MB | 11.6 MB |
| NSW, VIC, QLD, SA and WA suburbs | 50 | 1,316 s | 141 MB | 20.9 MB |
Plan on about 25 to 30 seconds and 0.3 to 0.75 MB of residential traffic per agent: the sites are read politely, one paced request stream per site, so a run's time grows with the number of agents, not with memory. The default 512 MB is enough for every run size above (the peak stayed under 150 MB), and the default Max results per run of 50 agents fits well inside the default run timeout. Western Australian agents use residential traffic too in a default run (their Domain and RateMyAgent profiles); only a datacenter-only run reads them from REIWA alone, with no residential traffic. The 20- and 50-agent times come from an earlier build that looked up licences between agents; the current build looks them up while the next agents are read, so these two rows are an upper bound.
What happens when a run reaches its time limit?
The actor watches the run's time limit and stops starting new agents early enough to score and deliver everything it already gathered: the reserve grows with every agent collected. The run then ends with a status that starts with Partial results: and says how many discovered agents were not processed. A run whose time limit is reached before a single agent could be collected delivers nothing, so it FAILS with a status that starts with Run time limit reached: and counts the agents and searches it could not process (also in OUTPUT.agentsNotProcessed and OUTPUT.searchesNotRun). The reserve kept for scoring and delivery is about 15 seconds plus a little per collected agent, and about 12 seconds more for each agent whose licence lookups are still running. Only delivered rows are billed, scores are computed over the agents that were collected (fewer peers than a complete run), no suburb leaderboard or network graph is built or billed from such an incomplete set, and incremental mode never marks an agent EXPIRED from such an incomplete scan, nor reports an agent as changed only because fewer peers were scored. Give the run a longer timeout or lower Max agents per suburb to cover everything.
How is a suburb name matched?
Each suburb you type is resolved once, from both directories' own suburb lists, and every source sweeps that same suburb, so one run never mixes two places of the same name. Names compare without case or punctuation, and St and Saint (and Mt and Mount) are the same word. An exact name wins. A city name with no suburb of exactly that name resolves to the city centre the directories list ("Darwin" is Darwin City NT 0800, "Hamilton" in New Zealand is Hamilton Central). The state, a postcode or the NZ region narrows the choice ("Richmond VIC", "Richmond 3121", "Ponsonby, Auckland", "Richmond (Nelson)", "Richmond Tasman"). A name several suburbs share, typed without a state, postcode or region ("Richmond", "Newtown", "St Kilda"), is never guessed: that suburb is not swept and not billed, and the run lists every candidate (for example Richmond VIC 3121; Richmond SA 5033; ...) in its status and in OUTPUT.suburbResolution, so you or your AI agent can pick one and run again. When a state is given but that state still has two suburbs of the name ("Richmond QLD"), the lowest postcode is swept and the warning names the other. A name no directory lists is reported as not found with the nearest names; a New Zealand suburb on an Australian run (or the other way round) says which country to set, and the country is never switched for you. A city centre that lists no agents of its own (Palmerston City NT 0830, whose agents are listed under Durack, Gray and the other suburbs of postcode 0830) is swept through the suburbs of its postcode, and a suburb that lists no agents at all is warned about in the log, because the name may be wrong.
Why did my run fail instead of returning an empty dataset?
If the sources refuse every request (suburb lists, agent name searches or the agent profiles they found), the run stops with a clear message so "no agents found" is never confused with "nothing could be read". A run whose only suburbs were ambiguous or not found fails too, naming the candidate suburbs (or the country to set), with nothing billed; so does a run that found no one in what it could search while some of its suburbs were ambiguous (the status names their candidates). When only some of them are refused, the run delivers what it read and its final status says how many profiles, name searches or suburb lists could not be read (nothing of them is billed). See the messages under Plan Requirement; the usual fix is a plan with proxy access and a residential connection in the target country under Connection. Run it again in a few minutes.
Can an AI agent use the output?
Yes, that is what the output is built for: see Use with AI agents above for a Python, a JavaScript and an MCP example with their real outputs, and for how to read matchEvidence, licenceCheck, licenceCandidates, score.drivers, unreadSections and soldHistoryComplete. The actor writes no summary itself; it returns structured records whose every value points back to its evidence, and your AI agent writes the answer. Every field is declared with a type and a description in the dataset schema, and the run summary OUTPUT carries the contract's schemaVersion.
- API or code: start a run and read the dataset items. View
defaultholds the agent reports,overviewevery row; filter onrecordType/changeTypefor other row kinds. - MCP: Apify's MCP server (mcp.apify.com) calls this actor with its
call-actortool and reads the rows withget-dataset-items. - Into your apps: the "Pipe results into your apps" field writes a condensed summary of each record (a title plus its key fields as plain text) into Notion, Linear or Airtable through your own authorised connectors; the full record stays in the dataset.
๐ Want more Realestate data?
Pair this actor with these related scrapers from the same team:
| ๐ Real Estate AU Extract detailed Australian real estate property listings with 30+ structured fields... | ๐ RealestateCo NZ Search and extract all real estate properties for sale, rent, and sold across New Zealand... |
| ๐ Real Estate AU Agents Scraper Scrape Australian real estate agent and agency profiles by suburb or profile URLs... | ๐ REIWA Scraper Scrape REIWA.com.au across sale, rental, sold, commercial, business and rural properties... |
| ๐ Homely.com.au Scrape Homely.com.au properties for sale, rent, and sold, plus agent finder, suburb... | ๐ HiPages Scraper Scrape business listings from HiPages Australia by category, location, or URL. Extract... |
๐ Browse all abotapi scrapers
๐ฌ Support
- ๐ Found a bug or a missing field? Open a ticket on the Issues tab. We usually reply within hours.
- ๐ ๏ธ Need another site, extra fields or a private build? Email contact@abotapi.com.
- โญ Enjoying it? A quick review on the actor page helps other users find it.