Given a school name + city/state, a Realtor.com school detail URL, or a property address, return the school's GreatSchools rating, parent-reviews summary, grades served, enrollment, student-teacher ratio, district, address, NCES code, and canonical URL. For property addresses, returns the list of assigned elementary / middle / high schools with each school's rating. This skill was captured from a live agent session on realtor.com and publishes here verbatim, exactly as an agent receives it.

Purpose

Return Realtor.com's school-detail payload for a given school — GreatSchools rating (1–10), parent-reviews summary (count + average), grades served, total enrollment, student-teacher ratio, district name + id, address, NCES code, GreatSchools id, catchment polygon (when present), and the canonical Realtor.com school-detail URL. Accepts three input shapes: a direct Realtor.com school URL, a school name + city/state, or a property address (which is resolved to its assigned elementary / middle / high schools). Read-only.

When to Use

A property-search agent needs school ratings to score listings.
A relocation agent comparing assigned-school quality across candidate addresses.
Bulk extraction of school metadata + catchment polygons across a metro.
Anywhere you'd otherwise scrape Realtor.com's school-detail HTML — the embedded __NEXT_DATA__ JSON is faster, structurally stable, and renders without JS.

Workflow

Realtor.com's school-detail pages ship a fully populated <script id="__NEXT_DATA__" type="application/json"> hydration blob inline in the HTML. No bot challenge fires on the school-detail surface — a single no-proxy browserless_agent call (goto + evaluate to pull and parse the blob in-page) is sufficient (~250–430 KB page, one round-trip). All required fields (rating, parent_rating, review_count, student_count, student_teacher_ratio, grades[], education_levels[], nces_code, greatschools_id, district.{id,name}, location.{...}, boundary GeoJSON catchment) live under props.pageProps.school in that blob. Lead with that plain goto+evaluate — the residential-proxy browser path is only needed for the property-address input, because property-detail pages sit behind Kasada Bot Defense (see Gotchas).

Step 1 — Normalize the input to a canonical school-detail URL

The canonical URL pattern is:

https://www.realtor.com/local/schools/{slug_id}

where {slug_id} = {Name-With-Dashes}-{schoolId} — e.g. Sylvia-Mendez-Elementary-078571861, Poway-High-School-078657741. schoolId is Realtor.com's internal id (9–10 digit numeric string, not the NCES id and not the GreatSchools id).

Input shape (a): direct school-detail URL → skip to Step 2.

Input shape (b): school name + city/state → resolve via the public autocomplete API (no auth, no anti-bot, no proxy):

GET https://parser-external.geo.moveaws.com/suggest
    ?input=<urlenc "<name> <city> <state>">
    &client_id=rdc-search-default
    &area_types=school
    &limit=10

These parser-external.geo.moveaws.com/suggest calls are plain HTTPS GETs — run them from any client. Only under restricted egress, route each through browserless_function: page.goto('https://parser-external.geo.moveaws.com/') first, then page.evaluate(async () => fetch('/suggest?...').then(r => r.json())) (the function runtime is a browser page context with no network egress until the page navigates to the origin).

Take the highest-scoring autocomplete[i] whose area_type === "school". The result includes slug_id, school_id, school, line (address), city, postal_code, state_code, centroid.{lat,lon}, and has_catchment (boolean — whether the school has a GeoJSON catchment polygon attached). Construct the URL as https://www.realtor.com/local/schools/{autocomplete[0].slug_id}.

If autocomplete is empty, retry once with area_types omitted (some out-of-database schools surface only when the type filter is dropped). If still empty, emit {success: false, reason: "school_not_found"}.

Input shape (c): property address → see "Property-address flow" below.

Step 2 — Open the school-detail page and pull the blob in-page

One browserless_agent call, no proxy, goto + evaluate:

{
  "commands": [
    {
      "method": "goto",
      "params": {
        "url": "https://www.realtor.com/local/schools/{slug_id}",
        "waitUntil": "load",
        "timeout": 45000
      }
    },
    {
      "method": "evaluate",
      "params": {
        "content": "(()=>{ const el = document.getElementById('__NEXT_DATA__'); if (!el) return JSON.stringify({ error: 'no_next_data' }); const school = JSON.parse(el.textContent).props.pageProps.school; return JSON.stringify(school); })()"
      }
    }
  ]
}

School-detail pages render the full hydration JSON without any proxy or anti-bot handling. The evaluate returns the parsed school object under .value (project it further in-page if it approaches the ~200k-char result cap — the boundary polygon can be large). If the goto lands on a Kasada interstitial instead (document.title / body would contain KPSDK / "reference ID"), retry once; if it recurs, fall back to a residential-proxy session (see Step 5).

Step 3 — Read the fields off the parsed blob

The evaluate above already returns props.pageProps.school. All fields live under school (see "Field map" below). The page also exposes props.pageProps.district (often null — district data is denormalized into school.district) and props.pageProps.nearbySchools (which despite the name is nearby cities/areas metadata, not nearby schools — do not use this for assigned-schools).

Step 4 — Emit the consolidated output

Map the parsed fields per the schemas in Expected Output. Critical mappings:

Output field	Source path in `__NEXT_DATA__`
`great_schools_rating`	`school.rating` (int 1–10, or `null` for private schools)
`parent_reviews.average`	`school.parent_rating` (int 1–5, or `null` if `review_count === 0`)
`parent_reviews.count`	`school.review_count`
`grades_served`	`school.grades` (array of strings like `["K","1",...,"5"]` — format for display as `"K-5"` if first=`"K"` and last=`numeric`, else join with commas)
`education_levels`	`school.education_levels` (array, e.g. `["elementary"]` or `["elementary","middle","high"]` for K-12)
`enrollment`	`school.student_count` (int)
`student_teacher_ratio`	`school.student_teacher_ratio` (float like `16.3`, or `null` for private schools — see gotcha)
`district`	`school.district.name` (string) — note `school.district.id` is a 11-char internal id, not the NCES district id
`address`	concatenate `school.location.{street, city, state, postal_code}`
`nces_id`	`school.nces_code` (string — sometimes 12 digits, sometimes 8 for older entries)
`great_schools_id`	`school.greatschools_id`
`funding_type`	`school.funding_type` ∈ `"public"`, `"private"`, `"charter"`
`url`	`https://www.realtor.com/local/schools/{school.slug_id}`
`catchment_polygon`	`school.boundary` (GeoJSON `MultiPolygon` — present only for public schools with `has_catchment: true`)

Step 5 — Browser fallback (only on Kasada wall)

If the plain no-proxy goto ever lands on a Kasada interstitial (a < 2 KB body containing KPSDK and "reference ID") on a school-detail URL — uncommon, but observed under aggressive batched requests — re-issue the same call with a residential proxy:

{
  "proxy": { "proxy": "residential", "proxyCountry": "us" },
  "commands": [
    {
      "method": "goto",
      "params": {
        "url": "https://www.realtor.com/local/schools/{slug_id}",
        "waitUntil": "load",
        "timeout": 45000
      }
    },
    {
      "method": "evaluate",
      "params": {
        "content": "(()=>{ const el = document.getElementById('__NEXT_DATA__'); return el ? el.textContent : null; })()"
      }
    }
  ]
}

The __NEXT_DATA__ blob renders identically here as on the plain path — parse it the same way. There's no release step — the session persists across calls keyed by proxy, so nothing needs tearing down.

Property-address flow (input shape c)

The Realtor.com property-detail page (/realestateandhomes-detail/...) is Kasada Bot Defense-protected and a proxy-less browserless_agent goto always lands on the interstitial. Two viable paths:

Path A — Catchment point-in-polygon (preferred; no property-detail page). Use this when the address has a known lat/lon and you only need elementary / middle / high assigned schools.

Resolve the address with the suggest API:

GET https://parser-external.geo.moveaws.com/suggest
    ?input=<urlenc address>
    &client_id=rdc-search-default
    &area_types=address
    &limit=5

Take the top result; record centroid.{lat,lon} and mpr_id.

Fetch the city's candidate schools via the suggest API (one call per education level you need):
```
GET https://parser-external.geo.moveaws.com/suggest
    ?input=<urlenc "elementary <city>">
    &client_id=rdc-search-default
    &area_types=school
    &limit=20
```
Filter results to those with has_catchment: true and same state_code as the address.
For each candidate, open its detail page with a no-proxy browserless_agent goto + evaluate (the Step 2 pattern) and read school.boundary (GeoJSON MultiPolygon) out of the returned blob. Run point-in-polygon (Shapely Point(lon, lat).within(shape(boundary)) or turf.booleanPointInPolygon) against the address centroid. You can also fold the point-in-polygon test into the in-page evaluate and return just a boolean + the school payload to keep results small.
Repeat for middle and high (use input=middle <city> / input=high <city>).
Emit one record per level with the matched school's full payload.

This path costs ~5–10 school-detail browserless_agent calls per address and completely avoids the property-detail page and Kasada.

Path B — Live browser to property page (fallback). Use when point-in-polygon is ambiguous (no candidate boundary contains the address — happens at district edges or in non-CA states with non-residential zoning), or when you need the exact "assigned schools" panel as Realtor.com renders it.

{
  "proxy": { "proxy": "residential", "proxyCountry": "us" },
  "commands": [
    {
      "method": "goto",
      "params": {
        "url": "https://www.realtor.com/realestateandhomes-detail/{address-slug}_{property-id}",
        "waitUntil": "load",
        "timeout": 45000
      }
    },
    { "method": "waitForTimeout", "params": { "time": 3000 } },
    {
      "method": "evaluate",
      "params": {
        "content": "(()=>{ const el = document.getElementById('__NEXT_DATA__'); if (!el) return null; const s = JSON.parse(el.textContent).props.pageProps.propertyDetail.schools || []; return JSON.stringify(s.filter(x => x.assigned === true)); })()"
      }
    }
  ]
}

The property page also ships a __NEXT_DATA__ blob — the evaluate reads props.pageProps.propertyDetail.schools[] (each item has school_id, slug_id, assigned: true|false, funding_type, rating, grades, education_levels, distance_in_miles) and filters on assigned === true for the catchment-assigned list. The assigned-schools panel is usually rendered without scrolling; if the array comes back empty, add a { "method": "scroll", "params": { "direction": "down" } } before the evaluate to force the "Schools" section to hydrate. The residential proxy arg is mandatory on this call — a proxy-less agent call gets Kasada-blocked on the first page load (Kasada isn't a solve-able type; the real browser + residential proxy clears it). Batching the whole flow in this one call's commands array is convenient; the session persists across calls keyed by proxy, so there's no release step — just repeat the same proxy on any follow-up call.

Site-Specific Gotchas

READ-ONLY. Never click "Save", "Contact agent", or any property-action button — the skill exists to extract data, not to interact.
The anti-bot is Kasada, not PerimeterX or DataDome. The interstitial fingerprints are <script>window.KPSDK={} and a request path of the form /{uuid}/{uuid}/ips.js?KP_UIDz=…. Don't waste time configuring DataDome cookie spoofing or PerimeterX header bypasses — they're the wrong vendor. A real browser driven through browserless_agent with a residential proxy (proxy: { proxy: "residential", proxyCountry: "us" }) is the only known consistent bypass for the protected surfaces — Kasada isn't a solve-able type, so it's the live JS runtime + residential IP that clears the challenge, not a captcha solver.
School-detail pages are NOT Kasada-protected (verified across 7 schools across multiple states + funding types, 2026-05-16). A plain no-proxy browserless_agent goto + evaluate returns the full __NEXT_DATA__ blob without any proxy or anti-bot handling — this is the canonical fast path for the school-detail and name-lookup input shapes. Treat a failure on a /local/schools/... URL as a transient hiccup, not a vendor change.
The /local/schools/search?searchTerm=... URL is a dead end. It resolves to Realtor.com's internal _error page (page /_error in __NEXT_DATA__, query {searchTerm, slugId: "search"}). Don't use it. Use the parser-external.geo.moveaws.com/suggest API instead.
props.pageProps.nearbySchools is a misnomer — its actual contents are nearby cities / neighborhoods / counties / zips metadata (slug_id, geo_statistics, recommended_cities, etc.), not other schools. Don't try to read assigned-schools from it.
Private schools have null ratings. GreatSchools doesn't rate private schools. For funding_type === "private", expect rating: null, student_teacher_ratio: null, district.name: null (but district.id is still populated with a synthetic state-prefix code like "06151428551"), and boundary: null (no catchment). Emit great_schools_rating: null + a null_rating_reason: "private_school_not_rated" flag rather than failing.
school.assigned is always null on the school-detail page. That field exists in the schema but is only populated when the school is referenced from a property-detail-page context. Don't read it from the school page.
school.boundary is a GeoJSON MultiPolygon. When present (public schools with has_catchment: true), it's a real catchment polygon usable for point-in-polygon assignment — see the Property-address Path A above. Polygons are sometimes 1000+ vertices; budget memory accordingly when iterating across a district.
nces_code length varies. Newer / mainstream entries are 12 digits (e.g. 060474000447); older charter / private entries are 8 (02061017). Both are valid — do not zero-pad or strip leading zeros.
Two different ids on the same school. school.id (and school.slug_id trailing segment) is Realtor.com's internal id (9-10 digits, e.g. 078571861 or 0772862241). school.greatschools_id is the upstream GreatSchools id (typically 7 digits, e.g. 0600034). school.nces_code is the federal id. The slug_id always uses school.id, not the GreatSchools id — never construct URLs with the GreatSchools id.
school.district.id is NOT the federal NCES district id. It's Realtor.com's internal id (11 chars, e.g. 06151428611). There is no NCES district code surfaced in the payload. If your output schema requires the federal district id, look it up separately.
school.grades is an array of strings, not a range. Public-school payloads use values like ["K", "1", "2", "3", "4", "5"]; preK is "PK". Render to "K-5" only when the array is contiguous; otherwise join with commas. Don't assume integer ordering — "K" and "PK" sort before "1" lexically only if you special-case them.
school.student_teacher_ratio is a float, not a colon string. Realtor.com returns 16.3; format to "16.3:1" only at the output layer.
parser-external.geo.moveaws.com/suggest query whitelist is strict. Accepted params: input, client_id, area_types, limit, include. postal_code, city, state, lat, lon, has_catchment are all rejected with whitelistValidation 400s. Filter / scope results client-side after the call.
The suggest API is unauthenticated and not rate-limited at typical agent volumes (tested at low double-digits RPS without throttling, 2026-05-16). It exposes school + address + city + street + county + zip area types. client_id=rdc-search-default is the Realtor.com web app's id; any non-empty value seems to work, but stick to rdc-search-default for forward-compat.
A raw (non-browser) HTTP fetch does not bypass Kasada, even through a residential proxy. Kasada requires JS execution to clear the interstitial; a plain fetch doesn't run JS regardless of proxy. For property-detail pages you need the real browser — a browserless_agent goto with a residential proxy — so the page's JS challenge actually runs. Verified 2026-05-16 that both proxied and unproxied raw fetches on /realestateandhomes-detail/... return the same Kasada interstitial.
Confirmed dead ends — don't re-probe:
- https://www.realtor.com/api/v1/schools/search → 404 ("Cannot GET").
- https://www.realtor.com/api/v1/hulk → 403.
- https://www.realtor.com/api/v1/rdc_search/schools → 404.
- parser-external.geo.moveaws.com/schools, /schools_search, /locality, /reverse_geocode → 404.
- m.realtor.com/... → 301 to www.realtor.com (no separate mobile surface).

Expected Output

Input shape (a) / (b) — school detail URL or name+city/state

{
  "success": true,
  "input_type": "school_detail_url",
  "school": {
    "name": "Sylvia Mendez Elementary",
    "school_id": "078571861",
    "slug_id": "Sylvia-Mendez-Elementary-078571861",
    "great_schools_id": "0600032",
    "nces_id": "060474000445",
    "funding_type": "public",
    "education_levels": ["elementary"],
    "grades_served": "K-5",
    "great_schools_rating": 7,
    "parent_reviews": { "count": 4, "average": 5 },
    "enrollment": 379,
    "student_teacher_ratio": "16.3:1",
    "district": {
      "name": "Berkeley Unified School District",
      "realtor_id": "06151428611"
    },
    "address": "2840 Ellsworth Street, Berkeley, CA 94705",
    "coordinate": { "lat": 37.857694, "lon": -122.262234 },
    "phone": "(510) 644-6290",
    "url": "https://www.realtor.com/local/schools/Sylvia-Mendez-Elementary-078571861",
    "has_catchment": true
  }
}

Private school (null rating)

{
  "success": true,
  "input_type": "school_name",
  "school": {
    "name": "Fairmont Private Schools - Historic Anaheim Campus",
    "school_id": "078696341",
    "funding_type": "private",
    "education_levels": ["elementary", "middle", "high"],
    "grades_served": "PK-12",
    "great_schools_rating": null,
    "null_rating_reason": "private_school_not_rated",
    "parent_reviews": { "count": 15, "average": 4 },
    "enrollment": 1835,
    "student_teacher_ratio": null,
    "district": { "name": null, "realtor_id": "06151428551" },
    "address": "...",
    "url": "https://www.realtor.com/local/schools/Fairmont-Private-Schools-Historic-Anaheim-Campus-078696341",
    "has_catchment": false
  }
}

Input shape (c) — property address

{
  "success": true,
  "input_type": "property_address",
  "property": {
    "address": "680 Grizzly Peak Blvd, Berkeley, CA 94708",
    "mpr_id": "1299668687",
    "coordinate": { "lat": 37.899275, "lon": -122.265644 },
    "url": "https://www.realtor.com/realestateandhomes-detail/680-Grizzly-Peak-Blvd_Berkeley_CA_94708_M12996-68687",
    "resolution_method": "catchment_point_in_polygon"
  },
  "assigned_schools": [
    {
      "level": "elementary",
      "name": "...",
      "school_id": "...",
      "great_schools_rating": 9,
      "grades_served": "K-5",
      "url": "..."
    },
    {
      "level": "middle",
      "name": "...",
      "school_id": "...",
      "great_schools_rating": 7,
      "grades_served": "6-8",
      "url": "..."
    },
    {
      "level": "high",
      "name": "...",
      "school_id": "...",
      "great_schools_rating": 8,
      "grades_served": "9-12",
      "url": "..."
    }
  ]
}

Failure shapes

// School name doesn't match anything in the suggest API
{ "success": false, "reason": "school_not_found", "input": "Foo Bar Academy Nowhere XX" }

// School-detail URL returned the Kasada interstitial AND browser fallback also failed
{ "success": false, "reason": "anti_bot_block", "vendor": "kasada", "evidence": "KPSDK present in 1.8KB response" }

// __NEXT_DATA__ block was missing or props.pageProps.school was empty
{ "success": false, "reason": "data_not_hydrated", "evidence": "no __NEXT_DATA__ script in response" }

// Property address: address resolves but no candidate school's boundary contains the point (district edge, non-residential parcel, or unincorporated area)
{ "success": false, "reason": "no_catchment_match", "address": "...", "candidates_checked": 12 }

// Property address: live-browser fallback blocked by Kasada despite Verified + proxies
{ "success": false, "reason": "anti_bot_block", "vendor": "kasada", "where": "property_detail_page" }

Realtor.com School Rating Lookup