OpenStreetMap Local Business Extractor — $2/1k places

Local-business records from OpenStreetMap, normalised into one flat row per business.

Price: $0.002 per business record returned — $2 per 1,000. Pay-per-event: a run that returns no usable record charges nothing, and the row cap you set is the ceiling on what you can spend.

Give it an area (a place name or a bounding box) and one or more OSM category tags. It runs a single Overpass query and returns clean rows — no HTML, no duplicated tag namespaces.

What one row looks like

{
  "name": "Tandartspraktijk Centrum",
  "category": "amenity=dentist",
  "lat": 52.3702,
  "lon": 4.8952,
  "address": "Damstraat 12, 1012 JM Amsterdam",
  "phone": "+31 20 555 0100",
  "website": "www.example.nl",
  "email": "",
  "opening_hours": "Mo-Fr 08:00-17:00",
  "osm_type": "node",
  "osm_id": 123456789,
  "permalink": "www.openstreetmap.org/node/123456789",
  "license": "ODbL-1.0",
  "attribution": "© OpenStreetMap contributors",
  "source": "OpenStreetMap via Overpass API"
}

phone prefers the bare phone tag and falls back to contact:phone; website and email behave the same way. address is assembled from the addr:* tags. website is reproduced exactly as OSM has it, normally including the scheme; the sample above is abbreviated. Records with no name tag are dropped rather than shipped as blanks.

Input

field	type	meaning
place	string	Area to search, resolved through Nominatim. Ignored when bbox is given.
bbox	[south, west, north, east]	Exact area in degrees.
categories	string list	OSM tags: amenity=dentist, shop=hairdresser, office=*.
maxResults	integer	Row cap, 1–1000.

Every run publishes its own fill rate

OSM coverage is uneven, so this actor measures itself instead of asking you to trust a claim. The last item in each dataset is a run_summary row:

{
  "record_type": "run_summary",
  "records": 23,
  "query_area": "Delft",
  "fill_rate": {
    "records": 23,
    "fields": {
      "phone":         {"tagged": 9,  "total": 23, "rate": 0.391},
      "website":       {"tagged": 11, "total": 23, "rate": 0.478},
      "email":         {"tagged": 3,  "total": 23, "rate": 0.13},
      "address":       {"tagged": 20, "total": 23, "rate": 0.87},
      "opening_hours": {"tagged": 12, "total": 23, "rate": 0.522}
    }
  }
}

Those numbers are illustrative of the shape, not of your area — the row you get back is computed from your own run's rows, field by field, and it is the number to judge the result by. The same object is printed to the run log as [fill_rate] {...}. It is not charged: only business records are billed.

Licence and attribution — please read

The data is OpenStreetMap, licensed ODbL 1.0 (www.openstreetmap.org/copyright). Every row carries license and attribution fields, and the actor refuses to write a row that does not. If you publish or redistribute this data, or anything derived from it, you must keep the attribution © OpenStreetMap contributors and honour the ODbL's share-alike terms. Do not strip those fields.

The code is CC0-1.0. The data is not — the licence travels with it.

Honest limitations

• OSM coverage varies by region and by category. In some cities most businesses are mapped with phone and opening hours; elsewhere you will get a name and a location and little else. This is thinner than a commercial POI database and is not claimed to be equivalent to one.
• Contact details are only what OSM already publishes. Nothing is enriched, guessed, or looked up on a second site. No email pattern generation.
• Empty strings mean "not present in OSM", not "does not exist".
• The public Overpass endpoint is a volunteer service. This actor issues one query per run, sequentially, with a descriptive User-Agent and exponential backoff on 429/504, and caps results at 1000 (operations.osmfoundation.org/policies/api). For large-scale extraction, run your own Overpass instance from a Geofabrik regional extract rather than increasing the run rate.
• Ways are returned by their centre point (out center), not their full geometry.