Childcare & Daycare Licensing Scraper

Scrapes state childcare/daycare licensing directories — provider name, address, license status, capacity, and (where the state publishes it) quality rating and deficiency counts.

Unlike nursing homes (one federal CMS dataset covers all states), childcare licensing has no national aggregator — every state runs its own system. That fragmentation is exactly why nobody's built this: it's a 50-state grind, not a hard technical problem. This actor uses a state-adapter architecture so coverage grows over time without changing how it's used — same state input either way, more states added as adapters are built.

Currently supported

State	Status	Providers	Data freshness	Notes
TX	✅ Live	14,368	Live/real-time	Real JSON API, no Cloudflare, no login
CA	✅ Live	39,184	Snapshot dated 2025-05-25	State open-data (CKAN) mirror — see caveat below

More states added over time. Request one if you need it.

No API key required

• TX: Texas Health & Human Services Child Care Regulation public search API (childcare.hhs.texas.gov) — a real POST/JSON endpoint behind an anonymous, auto-issued Bearer token.
• CA: California's own open-data portal (data.ca.gov), CKAN datastore_search API, combining the Child Care Centers and Family Child Care Homes resources.

No scraping, no browser, no Cloudflare fight for either state.

⚠️ California data freshness caveat

California's live facility-search tool (ccld.dss.ca.gov) was down at build time — the site itself displayed: "We are currently experiencing technical difficulties with the Facility Search and Download Data features... Please try your facility search and/or data download again later." Rather than build against a broken tool, this adapter uses California's official open-data mirror instead, which is unaffected by that outage.

The trade-off: that mirror is a periodic snapshot, not live. Every record's dataSnapshotDate field currently reads 2025-05-25 — over a year old as of this build — even though the CKAN catalog listing itself shows a recent "last touched" timestamp (only the catalog entry was re-indexed; the underlying data file was not refreshed). This is disclosed on every CA record via dataSnapshotDate, not hidden. If the live CCLD tool comes back up, a future version could switch to scraping it directly for current data.

Also unlike Texas: CA's open dataset does not include a star rating or deficiency count (CA does publish inspection reports, just not in this bulk file) — those fields are left blank on CA records rather than fabricated. Family Child Care Home addresses read "Unavailable" by law (a genuine legal restriction per the CCLD site's own FAQ, not a bug).

Use Cases

• Childcare franchise/PE roll-up site selection — map every licensed provider in a target market, filter by capacity
• M&A deal sourcing — find acquisition targets by provider type, license status, capacity
• Compliance/risk research (TX) — screen by deficiency count and license status
• Competitive market mapping — see the full licensed landscape in a region
• Historical/baseline research (CA) — the May 2025 snapshot is still useful for market-sizing and baseline mapping even though it isn't current

Input

Field	Type	Description
state	String	Required. TX (live) or CA (2025-05-25 snapshot).
providerName	String	Optional — only return providers whose name contains this text (server-side filter on both states, verified working).
maxResults	Integer	Maximum providers to return. 0 = no limit (default). Full TX ≈ 14,368; full CA ≈ 39,184.

Note: there is no location/city filter in this version. Texas's underlying API's location search requires a resolved lat/long + radius (from a client-side geocoder), not a plain city string — a raw city parameter was tested and confirmed to have no effect on TX (it silently returned the full unfiltered statewide set), so it isn't exposed rather than shipping a filter that quietly does nothing. Pull the full state and filter client-side for now.

Output

Field	Description
providerId	State-assigned provider/facility identifier
providerName	Facility or home name
providerType	e.g. Licensed Center (TX) / Day Care Center, Family Day Care Home (CA)
address, city, state, zipCode, county	Location (CA family-home addresses may read "Unavailable" by law)
licenseStatus	e.g. Full Permit (TX) / Licensed, Closed (CA)
licenseIssueDate	Date of current license
closedDate	Closure date, if closed (CA)
capacity	Total licensed capacity
starRating	State quality rating — TX only, blank for CA
deficiencyCount	Citations/deficiencies on record — TX only, blank for CA
agesServed	Age groups served — TX only
acceptsSubsidy	Accepts subsidized/voucher care — TX only
licensee	Licensed owner/operator of record — CA only
dataSnapshotDate	Snapshot date for non-live states (CA); blank for live states (TX)
phone, email	Contact info, where available
scrapedAt	Timestamp when the record was collected

Example Input

All Texas providers (live data):

{ "state": "TX" }

All California providers (2025-05-25 snapshot):

{ "state": "CA" }

Providers matching a name in either state (e.g. a chain/brand audit):

{ "state": "CA", "providerName": "KinderCare" }

Notes

• Data sources: Texas Child Care Search (live), California Community Care Licensing Facilities (open-data snapshot).
• No API key, no rate limit encountered in testing on either state.
• Real measured runtimes on the Apify platform (not estimates, not local-only numbers):
  • Full TX run (14,368 providers): ~4 min — each 2000-row page costs a flat ~30s server-side on Texas's end, independent of offset.
  • Full CA run (39,184 providers, both resources): ~8.5 min — each 5000-row page costs a flat ~70–76s, independent of offset or which resource.
  • Both states show the same pattern: flat per-request latency regardless of payload size or pagination offset, and platform runtimes are 5–8x slower than local runtimes for the same request pattern (CA: 93s locally vs. ~8.5 min on-platform for the identical run). This points to something in Apify's outbound network path (very possibly Cloudflare treating Apify's datacenter IP ranges differently than a residential/local IP) rather than a bug in this actor's code — worth the same check for any future state adapter before assuming local timing will hold on the platform.