- Home
- Skills
- Sales & Lead Gen
- Local Lead Finder (OpenStreetMap)
Local Lead Finder (OpenStreetMap)
For **freelancers and small agencies who sell websites, online booking or local SEO to local businesses**, anywhere in the world.
$2
Works with the AI Agents you already use
Local Lead Finder (OpenStreetMap)
Example session with this skill installed
Find hair salons, barbers and beauty salons in Ghent, Belgium. I build websites for small businesses. Who should I look at first?
- Read your context and instructions
- Compiled the local lead finder
Real output, 23 Oct 2026, from two live runs of the script. Business names are replaced with generic ones, phone digits are masked and street names are removed. Everything else is as returned.
Area: Gent, Oost-Vlaanderen, Vlaanderen, België (boundary relation 897671, bbox ~439 km2)
Overpass: overpass-api.de → HTTP 502
Overpass: overpass.private.coffee → unreachable (read timed out)
Overpass: maps.mail.ru → OK
# Local leads: hairdresser, barber, beauty in Ghent, Belgium
247 leads after filtering (264 objects found, 17 excluded).
Niche: hairdresser 205 · barber 9 · beauty 42
Lead type: no_website 172 · social_only 8 · booking_page_only 6 · has_website 61
Priority: A 49 · B 32 · C 166 (a sorting aid, not a probability).
Data coverage in OSM: phone 30% · website 27% · opening hours 32% · street address 53% · email 13%.
Excluded: no name in OSM 11 · chain (brand:wikidata) 4 · brand tag 1 · duplicate 1.
| Pri | Score | Lead type | Name | Phone | Next step |
|-----|-------|-------------------|----------------|------------------|-----------------------------------|
| A | 100 | social_only | Hair salon A | +32 9 ••• •• •• | website pitch (only Facebook) |
| A | 95 | no_website | Hair salon B | +32 470 •• •• •• | website pitch, check search first |
| A | 90 | social_only | Hair salon C | +32 487 •• •• •• | website pitch (Facebook + Insta) |
| A | 80 | has_website | Hair salon E | +32 9 ••• •• •• | run local-business-site-audit |
| A | 75 | booking_page_only | Beauty salon F | … | own site / SEO (Treatwell only) |
Data (c) OpenStreetMap contributors, available under the ODbL - https://www.openstreetmap.org/copyright
Second live run, same day: car repair and tyre shops in Braga, Portugal. Three public servers were busy (HTTP 502, a timeout, HTTP 504); the script waited 20 s and got the answer on its second round. 66 objects →
45 leads; excluded 13 unnamed places and
8 chains (Midas, Norauto, Carglass, Bosch Car Service and others). OSM coverage was thin and the summary says so: phone 18 %, website 7 %, and 36 of 45 leads had no contact in OSM, so they were ranked C. The top lead: an independent garage whose listed website is a negocio.site page Google switched off in 2024 — flagged as dead_google_site with a website pitch as the next step.
Connects securely to your tools. The creator never sees your data.
About this skill
For freelancers and small agencies who sell websites, online booking or local SEO to local businesses, anywhere in the world. The first job in prospecting is a list of real, independent businesses in one area, along with what each of them is missing. This skill builds that list from OpenStreetMap.
Ask in plain words:
- "Find hair salons, barbers and beauty salons in Ghent."
- "Car repair and tyre shops in Braga, Portugal. I sell websites."
- "Dentists in Kreuzberg, Berlin. Which ones have no website?"
The agent runs the included script (scripts/find_leads.py, Python standard library only):
- Finds the area with one OpenStreetMap geocoder request and checks that it picked the right place.
- Builds one small Overpass query from a niche → OSM tag map with 46 niches, each tag checked against its OSM wiki page and taginfo. Aliases like "mechanic" or "hair salons" work, and
--tag key=valuecovers anything else. - Sends it politely: identifying User-Agent, timeouts, moves to the next of 3 public Overpass instances when one is busy, waits 30 s on a rate limit as the usage policy asks, at most 6 attempts, no parallel requests or tile loops. The raw answer is saved for offline re-filtering.
- Filters out businesses that won't buy from you: chains and franchises (
brand:wikidata/brand/operator:wikidata, a configurable chain-name list, names repeated 3+ times), public institutions, closed, unnamed and duplicate places. Everything excluded goes to a separate file with the reason. - Normalises phones to international format for 70+ countries, keeping the original and flagging numbers it can't trust.
- Scores every lead and tells you what to pitch:
no_website(none listed in OSM — confirm with a quick search),social_only,dead_google_site(business.site / negocio.site pages Google switched off in 2024),booking_page_only(Booksy, Fresha, Treatwell…),has_website(run the companion Local Business Website Audit skill first),weak_data.
You get leads.csv (priority A/B/C, score, lead type, name, niche, address, phone, email, website, social links, hours, flags, next step, OSM link), leads.md (counts per niche and lead type, OSM data completeness in that area, exclusions, top 10, attribution) and leads_excluded.csv.
Honest by design: it says "no website listed in OpenStreetMap", never "has no website". The score is a sorting aid, not a sales probability. Thin coverage is reported as thin. It never contacts a business, never visits their websites and never scrapes Google Maps or any site that forbids it. Data © OpenStreetMap contributors (ODbL), attribution included in every output.
- OpenStreetMap coverage varies a lot by city and niche. Many businesses have no phone, website or opening hours in OSM, and some aren't mapped at all.
- "No website" means none listed in OSM, so verify before pitching. Data can also be stale (closed or moved businesses).
- Chain detection relies on OSM brand tags, a built-in name list you can extend, and repeated names. Some small local chains will slip through, and the occasional independent business may be excluded — review the excluded file.
- One area per run, up to about 2,500 km² (bounding box), on purpose. It is not built for country-wide or bulk scraping.
- The public Overpass servers are often busy and can refuse requests. The skill doesn't work around their limits; for regular commercial use the operators ask you to use your own or a paid instance (--endpoint).
- It does not check the businesses' websites, Google Business Profiles, reviews or rankings. Use the companion audit skill for that.
- Not legal advice on data protection or marketing rules.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 25 days ago
Requires
Permissions
Allowed Hosts
- Network: one request to OpenStreetMap's Nominatim geocoder to find the city boundary (skipped if you give a bounding box or relation id), and one Overpass API query per run; up to 6 attempts across 3 public instances if servers are busy. It identifies itself with a User-Agent and follows the Overpass and Nominatim usage policies. - It does not visit any business's website and does not contact, email or message anyone. No Google Maps scraping. - Files: it writes <out>.csv, <out>_excluded.csv, <out>.md and <out>.overpass.json in the folder you choose. Nothing else is stored. - No account, API key, token or required environment variables.