US Building Permits — 19 sources, one schema

Building permits from nineteen US municipal open-data portals, normalised into a single schema with one date format, one address field, and one currency field.

~11,150,026 permit records across 19 actively-maintained sources.

Source	State	Records	Last verified
New York City	NY	~3,989,647	2026-07-28
Austin (all trades)	TX	~2,367,694	2026-07-28
Chicago	IL	~842,592	2026-07-28
New York City (electrical)	NY	~580,864	2026-07-30
Seattle (electrical)	WA	~498,027	2026-07-28
New Orleans	LA	~460,490	2026-07-29
Los Angeles	CA	~403,243	2026-07-25
Montgomery County	MD	~319,392	2026-07-28
Austin	TX	~285,851	2026-07-06
Seattle (trade)	WA	~284,688	2026-07-29
Seattle	WA	~191,830	—
Montgomery County (residential)	MD	~185,696	2026-07-30
Cincinnati	OH	~178,460	—
Montgomery County (mechanical)	MD	~157,388	2026-07-30
Mesa	AZ	~155,569	2026-07-27
Baton Rouge	LA	~142,483	—
Marin County	CA	~50,280	2026-07-29
Montgomery County (commercial)	MD	~41,675	2026-07-30
Cambridge	MA	~14,157	2026-07-28

Registered but excluded by default (abandoned upstream, kept so the staleness is documented rather than rediscovered):

• Dallas, TX — last record 2020-08-29
• Gainesville, FL — last record 2023-02-28

Why this exists

Every city below publishes permits on Socrata, so the transport is identical and trivial. Everything else differs. The issue-date column alone is spelled nine ways:

City	Column	Type
Chicago	issue_date	timestamp
New York City	issuance_date	text MM/DD/YYYY
Seattle	issueddate	timestamp
Austin	issue_date	timestamp
Los Angeles	issue_date	timestamp
Cincinnati	issueddate	timestamp
Baton Rouge	issueddate	timestamp
Mesa	opened_date	timestamp (application date)
Cambridge	issue_date	timestamp

Addresses are worse — Chicago splits into number/direction/name, NYC into house/street/borough, Seattle keeps one preformatted line, Cambridge uses full_address. Project value appears as estprojectcost, valuation, total_job_valuation, projectvalue, building_cost, and icc_value.

Writing nineteen adapters is a week. Maintaining them as cities silently rename and retype columns is the actual cost, and it never ends.

* NYC's dataset holds 4M rows going back to 1989, but only ~4.6k permits were issued in 2026 and ~10k in 2025 — the bulk is historical.

Output schema

{
  "source_city": "chicago",
  "city_label": "Chicago",
  "state": "IL",
  "permit_number": "101084879",
  "permit_type": "PERMIT - RENOVATION/ALTERATION",
  "permit_subtype": "STANDARD PLAN REVIEW",
  "status": "ACTIVE",
  "description": "INTERIOR ALTERATION OF EXISTING CORRIDORS AND RESTROOMS...",
  "issue_date": "2026-07-28",
  "issue_date_is_approximate": false,
  "application_date": "2026-06-25",
  "expiration_date": null,
  "address": "600 W CHICAGO AVE",
  "city_name": "Chicago",
  "zip_code": "60654",
  "latitude": 41.8965851912,
  "longitude": -87.6429961817,
  "project_value_usd": 351750.0,
  "fee_usd": 600.0,
  "contractor_name": "MATTHEW RYAN",
  "applicant_name": null,
  "dwelling_units": null,
  "source_url": "https://data.cityofchicago.org/resource/ydr8-5enu.json",
  "source_dataset_id": "ydr8-5enu"
}

permit_number, issue_date, address, latitude and longitude are populated for 100% of records across all nine live cities.

Honest field-availability notes

Some fields are absent because the city does not publish them, not because normalisation failed. Rather than emit a plausible-looking null and let you discover this in your own funnel metrics:

• No project valuation: NYC, Mesa
• No contractor name: Seattle, Austin, Los Angeles
• No site ZIP: Mesa, Cambridge (Mesa publishes contractor ZIP only)
• issue_date_is_approximate: true on Mesa — it publishes an application date, not an issue date. Every record is flagged so you can exclude it.

Set includeRawRecord: true to attach the untouched municipal record under raw and check anything yourself.

Who this is for

• Solar / roofing / HVAC contractors buying geocoded permit leads. LA additionally exposes a solar flag.
• Construction-materials suppliers sizing demand by metro and value band.
• Proptech and lead-gen platforms needing a permit feed without owning nine scrapers.
• Market researchers tracking construction activity across metros.

Usage

Typical lead-generation run — high-value residential work in the last month:

{
  "cities": ["chicago", "los_angeles", "seattle", "cincinnati"],
  "issuedAfter": "2026-07-01",
  "minProjectValue": 50000,
  "requireAddress": true,
  "maxResults": 5000
}

Solar-specific:

{ "cities": ["los_angeles", "austin", "mesa"], "search": "solar", "issuedAfter": "2026-01-01" }

Careful with minProjectValue: records where the city publishes no valuation are excluded, so setting it drops NYC and Mesa entirely. That is deliberate — a lead you cannot size did not meet your threshold.

Local development

No Apify runtime needed; the scraping logic is independent of the platform.

pip install requests
python local_run.py --coverage                       # registry + row counts
python local_run.py --cities chicago,seattle --issued-after 2026-06-01 --limit 20
python verify.py                                     # health-check every source

verify.py is the maintenance early-warning system. Run it on a schedule — it catches the three ways these feeds break:

1. Abandoned upstream (Dallas: metadata claims fresh, data ends 2020)
2. Renamed column (mapping is alias-tolerant, but a rename matching no alias silently nulls a field)
3. Retyped column (NYC's text dates make range filters return plausible-looking nonsense instead of an error)

It caught a real bug during development: a suffix column on Chicago that does not exist.

Legal

Public, logged-out municipal open-data endpoints served through each city's official Socrata API. No authentication, no account, no terms accepted, no anti-bot circumvention, no personal contact data. Contractor and applicant names are published by the municipality as part of the public record.

Rate limits are respected with exponential backoff — these are public-good servers. Supplying a free Socrata app token raises your limit above the shared anonymous pool.