Geocoding Scraper — Address ↔ Coordinates (OpenStreetMap)

Key-less geocoding on OpenStreetMap data for agents, CRMs and data pipelines. Forward (address or place → lat/lon + full address breakdown), structured queries (street/city/postcode fields for messy CRM data), reverse (lat,lon → address at the zoom level you need), OSM id lookup, bounded POI search inside a box, boundary polygons as GeoJSON and names in any language. Two engines: Nominatim (exact, rich) and Photon (typo-tolerant). Worldwide, no API key, no proxy.

Need it tailored to your workflow? Missing a field, a filter, another source or a different output format? Open an issue and describe what you need — we read every request and adapt the actor.

Use cases

• CRM address cleaning — geocode 500 customer addresses with structuredQueries, keep found, lat/lon, normalized postcode/city; with emitNotFound: true rows stay 1:1 with input.
• Store / branch enrichment — reverseZoom: 10 turns delivery GPS points into city names for territory reports.
• Local lead lists — queries: ["pharmacy"] + viewbox + bounded: true lists every mapped pharmacy in a district with street and house number; extraTags adds opening hours and websites where mapped.
• Service-area maps — polygon: true gives district/city boundaries as GeoJSON for coverage checks.
• Multilingual labels — lang: "ru" or nameDetails: true for Cyrillic / Kazakh / local names on the same object.
• Dedupe & entity resolution — stable osmType+osmId, placeRank, importance and bestMatch to pick the right candidate among ambiguous names.

Input

Field	Type	Default	Notes
queries	string[]	—	Text address/place → forward; lat,lon → reverse; R62422 / N1,W2 → lookup. Points are also read in other notations: 40.7488° N, 73.9854° W, 48°51'30"N 2°17'40"E, (51.5007, -0.1246), 48,8584; 2,2945, lat: 51.5 lon: -0.12, POINT(2.2945 48.8584) (lon lat) and Google Maps / openstreetmap.org links; the log and SUMMARY.readAsCoordinates show what each was read as.
structuredQueries	object[]	—	{ amenity, street, city, county, state, country, postalcode, label } (Nominatim).
mode	select	auto	auto, forward, reverse, lookup.
engine	select	nominatim	nominatim or photon (structured & lookup always use Nominatim).
lang	string	en	Accept-Language, e.g. ru, kk, de, ru,en; empty = en. Photon: en/de/fr only.
countryCodes	string	—	kz,ru — ISO-3166 alpha-2, separated by commas (or ;, `
viewbox	string	—	minLon,minLat,maxLon,maxLat, e.g. Almaty 76.75,43.15,77.10,43.40 (spaces, or ; with decimal commas, work too).
bounded	bool	false	Only results inside viewbox (enables area POI search).
featureType	select	—	country, state, city, settlement.
layers	multi	—	address, poi, railway, natural, manmade.
limit	int	1	Results per query, 1–50. Nominatim returns at most 40 (a higher value is lowered to 40 with a warning), Photon at most 50.
reverseZoom	int	18	18 building · 17 street · 16 major street · 14 suburb · 12 town · 10 city · 8 county · 5 state · 3 country.
addressDetails	bool	true	address object + derived columns.
extraTags	bool	false	extraTags: website, opening_hours, wikidata, population…
nameDetails	bool	false	nameDetails: name:ru, name:kk, official_name…
polygon	bool	false	geojson outline.
polygonThreshold	number	0	Simplification tolerance in degrees (0.005 ≈ city level).
dedupe	bool	true	Nominatim result dedupe.
emitNotFound	bool	false	Emit found: false rows to keep 1:1 alignment. Such rows are charged like hits, so this is opt-in; not-found queries are always listed free in SUMMARY.notFoundInputs.
outputFields	string[]	—	Projection in your order, e.g. ["query","lat","lon","countryCode"]; query is always included. Wrong case, aliases (lng, zip) and near misses are corrected with a warning; unknown names are ignored with a warning; a list with no known column fails the run and lists the columns.
maxItems	int	10000	Hard cap on rows (one entry gives up to limit rows).
osmTag	string	—	Photon: amenity:cafe, shop, !amenity:parking.
photonLayers	multi	—	Photon: house, street, locality, district, city, county, state, country, other.
biasLat, biasLon	number	—	Photon: prefer results near this point.
nominatimUrl	string	—	Base URL of your own Nominatim server (the part before /search). No 1.1 s pace and no per-run cap there. Empty = nominatim.openstreetmap.org.
photonUrl	string	—	Base URL of your own Photon server (the part before /api). Empty = photon.komoot.io.

Off-list values of the select fields (mode, engine, featureType, layers, photonLayers) are refused by Apify before the run starts; the allowed values are the ones listed above.

Reference

Nominatim vs Photon

	Nominatim	Photon
Matching	exact tokens, structured fields	fuzzy / prefix (typos OK)
Address breakdown	full (address object, 20+ keys)	street, housenumber, district, city, state, postcode, country
Reverse	yes, zoom-controlled	yes (nearest object)
OSM id lookup, polygons, extra tags, name details	yes	no
Language	any (accept-language)	en / de / fr
POI filter	layers, featureType, bounded box	osm_tag, layer, bbox, location bias
Policy	max 1 request/s, one at a time, identify the app, cache results (the actor keeps the pace, sends one request at a time with its own User-Agent, and requests a repeated entry of a run once; it keeps nothing between runs)	fair use (the actor uses the same pace)

reverseZoom levels

zoom	returns
3	country
5	state
8	county
10	city
12	town / borough
13	village / suburb
14	neighbourhood
16	major streets
17	major and minor streets
18	building

Common type / category values

boundary/administrative (countries, regions, cities, districts) · place/city|town|village|suburb|neighbourhood · highway/residential|primary|… (streets) · building/yes|office|apartments|… · amenity/pharmacy|cafe|bank|school|… · shop/* · tourism/* · office/*. Full tag catalogue: https://wiki.openstreetmap.org/wiki/Map_features

Examples

Batch-geocode CRM addresses (1:1 rows)

{
  "structuredQueries": [
    { "street": "Nauryzbai Batyr 154A", "city": "Almaty", "country": "Kazakhstan", "label": "Kaspi HQ" },
    { "street": "Mangilik El 55", "city": "Astana", "country": "Kazakhstan", "label": "Client 2" }
  ],
  "lang": "ru",
  "emitNotFound": true,
  "outputFields": ["query", "found", "lat", "lon", "displayName", "postcode", "city", "osmUrl"]
}

All pharmacies in a district (POI lead list)

{ "queries": ["pharmacy"], "viewbox": "76.85,43.20,77.00,43.30", "bounded": true, "limit": 40, "extraTags": true, "lang": "ru" }

GPS points → city names for reporting

{ "queries": ["43.2363,76.9457", "51.1282,71.4307", "42.3417,69.5901"], "reverseZoom": 10, "outputFields": ["query", "city", "state", "countryCode"] }

City boundaries as GeoJSON

{ "queries": ["Almaty", "Astana"], "featureType": "city", "countryCodes": "kz", "polygon": true, "polygonThreshold": 0.005 }

Fuzzy search of user-typed input (Photon)

{ "queries": ["Almty Nauryzbai Batyr 154"], "engine": "photon", "limit": 3, "biasLat": 43.24, "biasLon": 76.95 }

Output

{
  "query": "Nauryzbai Batyr 154A, Almaty",
  "mode": "forward",
  "engine": "nominatim",
  "rank": 1,
  "bestMatch": true,
  "found": true,
  "placeId": 213728564,
  "osmType": "way",
  "osmId": 51267846,
  "osmUrl": "https://www.openstreetmap.org/way/51267846",
  "displayName": "Kaspi.kz, 154А, Nauryzbai Batyr Street, Керемет, Көктем, Bostandıq District, Almaty, 050013, Kazakhstan",
  "name": "Kaspi.kz",
  "lat": 43.2399856,
  "lon": 76.9388247,
  "category": "building",
  "type": "office",
  "addressType": "building",
  "placeRank": 30,
  "importance": 0.0000743,
  "boundingBox": [43.2398836, 43.2400877, 76.9383802, 76.9392699],
  "country": "Kazakhstan",
  "countryCode": "KZ",
  "state": null,
  "county": null,
  "city": "Almaty",
  "district": "Bostandıq District",
  "suburb": "Көктем",
  "postcode": "050013",
  "road": "Nauryzbai Batyr Street",
  "houseNumber": "154А",
  "address": { "building": "Kaspi.kz", "house_number": "154А", "road": "Nauryzbai Batyr Street", "city": "Almaty", "ISO3166-2-lvl4": "KZ-75", "postcode": "050013", "country": "Kazakhstan", "country_code": "kz" },
  "extraTags": { "building:levels": "7" },
  "nameDetails": null,
  "geojson": null,
  "licence": "Data © OpenStreetMap contributors, ODbL 1.0. https://osm.org/copyright",
  "sourceUrl": "https://nominatim.openstreetmap.org/search?q=…",
  "fetchedAt": "2026-09-12T23:50:01.000Z"
}

Field	Description
query, mode, engine	Input label (or structured label), detected mode, engine used.
rank, bestMatch, found	1-based rank; rank 1 is bestMatch; found: false rows carry nulls.
placeId, osmType, osmId, osmUrl	Stable OSM identity (use osmType+osmId as key; placeId changes between Nominatim imports).
displayName, name	Full label and object name.
lat, lon, boundingBox	WGS84; bbox as [minLat, maxLat, minLon, maxLon] (Nominatim order).
category, type, addressType, placeRank, importance	OSM class/type, address level, 0–1 relevance.
country…houseNumber	Normalized address columns; raw parts in address.
extraTags, nameDetails, geojson	Only when requested.
structured	Echo of the structured query object (structured rows only).

The dataset schema describes every column (type, title, description, example) for agents and MCP clients.

SUMMARY key-value record: { queries, found, notFound, rows, errors[{query, error}], engine, errorCount, notFoundCount, notFoundInputs[], notFoundRowsEmitted, notSentOverCap, notSentRefused, notProcessed, notProcessedInputs[], requests{server: n}, reusedAnswers, refusedBy, warnings[], delivered, stoppedBy, status } — notProcessedInputs lists (up to 100) entries the run did not reach (spending limit, timeout or maxItems), so you can run just those again.

Use it from code / agents

curl -X POST "https://api.apify.com/v2/acts/yadroo~osm-geocode/run-sync-get-dataset-items?token=$APIFY_TOKEN" \
  -H 'Content-Type: application/json' -d '{"queries":["Almaty, Kazakhstan","43.2363,76.9457"],"lang":"ru"}'

import { ApifyClient } from 'apify-client';
const client = new ApifyClient({ token: process.env.APIFY_TOKEN });
const run = await client.actor('yadroo/osm-geocode').call({ structuredQueries: [{ street: 'Nauryzbai Batyr 154A', city: 'Almaty', country: 'Kazakhstan' }] });
const { items } = await client.dataset(run.defaultDatasetId).listItems();

from apify_client import ApifyClient
client = ApifyClient(os.environ["APIFY_TOKEN"])
run = client.actor("yadroo/osm-geocode").call(run_input={"queries": ["pharmacy"], "viewbox": "76.85,43.20,77.00,43.30", "bounded": True, "limit": 40})
items = client.dataset(run["defaultDatasetId"]).list_items().items

MCP: connect https://mcp.apify.com and call the yadroo/osm-geocode tool with the same JSON.

Pricing

Pay per event: $0.005 per run start + $0.002 per result row (from 5 November 2026; until then $0.001 + $0.001) (not-found rows are included only when you turn emitNotFound on; failed queries and entries that were not sent are never charged). Bronze −5 %, Silver and above −10 % on the row price (from 5 November 2026; until then −10 / −20 / −30 %); the start event is the same on every plan; platform usage is included. From 5 November: 100 addresses at limit: 1 ≈ $0.21; a bounded POI search returning 40 rows ≈ $0.085.

The run respects your maximum cost per run: it stops when the next row would go over it and ends with "Stopped at your spending limit: N rows delivered".

Limits & FAQ

• Usage policy and speed. The public Nominatim server allows at most 1 request per second, from one machine, one request at a time, with no parallel or distributed jobs, and asks bulk users to cache results; Photon asks for fair use. The actor starts one request every 1.1 s per server, requests a repeated entry only once per run (repeats reuse the first answer while the kept answers fit in 16 MB, beyond that a repeat is asked again; nothing is kept between runs), sends at most 1,000 requests per run to each public server (a migrated or restarted run continues the count), and slows down when a server throttles: after a 429 (too many requests from this Apify server — other users share its IP), a 5xx or a timeout it pauses 30 s, 60 s, then 120 s and continues at half the pace each time (down to one request per 4.4 s); the log says so and SUMMARY.pausesAfterThrottling counts it. It stops sending to a server that answers 401/403, that keeps throttling after these pauses, or that refuses from the very first request (then after one 20 s pause, so the run ends within a minute) — those entries are listed in SUMMARY.errors as not sent, never charged. Do not start several runs in parallel against the public servers. For big batches, run your own Nominatim or Photon and set nominatimUrl / photonUrl: no pace and no per-run cap there.
• Capacity per run. The default 20-minute timeout fits about 1,000 entries — the per-run cap — at 1.1 s each while the server does not throttle. When a list is longer, the log says so at the start (how many fit and the timeout that would fit them), and a bigger batch logs a progress line every 15 s (entries done, rows saved, time left) — the run is working, not stuck. The run stops starting new requests shortly before its timeout, keeps every row it has, ends SUCCEEDED with "Stopped before the run timeout: N rows saved" and lists the entries it did not reach in SUMMARY.notProcessedInputs. A longer list: raise the timeout (up to the 1,000-request cap), or run the rest later, one run after another.
• Accuracy. OSM coverage varies: exact house numbers exist for most cities in Europe/CIS/US; rural addresses may resolve to the street or village (check addressType/placeRank).
• Ambiguity. Use countryCodes, viewbox, featureType or structuredQueries; inspect importance and rank when limit > 1.
• Errors. HTTP 429/5xx and timeouts → up to 3 attempts with backoff; a failing query is listed in SUMMARY.errors (no row, not charged) and the run continues. This also covers queries that do not fit a forced mode (e.g. an address while mode: "reverse") and structured queries with no address part (only label): they are reported per query, the rest of the batch still runs. Nothing fetched at all (for example the public server refuses this Apify server from the first request) → the run fails within about a minute, uncharged, and the status message gives the reason once with what to do.
• Licence. Data © OpenStreetMap contributors, ODbL — attribution required when you display it; Nominatim results may not be used for autocomplete-as-you-type (use Photon for that).
• Restarts. If Apify migrates or restarts the run, it continues where it stopped; rows already saved are not saved or charged again.
• Roadmap. Distance/nearest computation between rows.

---

Made by Yadroo · Sibling actors: open-meteo-weather · ip-intel · public-holidays · krisha-kz