Snow Leo Data — MCP servers
Five remote MCP servers that give an AI agent 21 web-data tools: local business
leads, job listings, marketplace catalogues, compliance records and public media
feeds. Every tool is a published Apify Actor, exposed over the Model Context
Protocol by Apify's own hosted gateway at mcp.apify.com.
- Transport: Streamable HTTP. No install, no npm package, no local process.
- Auth: OAuth with your own Apify account (or a Bearer token).
- Billing: you pay Apify directly, per result. Prices are listed below.
The servers are split by subject on purpose. An agent handed twenty-one tools at once picks badly; an agent handed four tools that all answer one kind of question picks well. Connect only the server you need.
The five servers
1. Leads & Contacts
https://mcp.apify.com/?actors=snow_leo_data/google-maps-scraper,snow_leo_data/yellow-pages-scraper,snow_leo_data/duckduckgo-scraper,snow_leo_data/openstreetmap-business-scraper,snow_leo_data/website-contact-scraper
Five different indexes of the same world — Google's, Yelp/Apple's, the printed directories, the open map, and the company's own website — so an agent that comes up short in one can try another.
2. Jobs & Hiring
https://mcp.apify.com/?actors=snow_leo_data/ats-jobs-scraper,snow_leo_data/seek-scraper,snow_leo_data/jobs-ch-scraper,snow_leo_data/the-muse-jobs-scraper
One tool for jobs straight from the employer's own ATS, and three for the job boards that dominate a region. All four return the same kind of row, so an agent can compare a Swiss board against a US careers page without reshaping anything.
3. Marketplaces & Prices
https://mcp.apify.com/?actors=snow_leo_data/amazon-product-scraper,snow_leo_data/shopify-inventory-scraper,snow_leo_data/app-store-scraper,snow_leo_data/google-play-scraper,snow_leo_data/airbnb-scraper,snow_leo_data/redfin-scraper
Every tool here returns a priced listing out of somebody's catalogue — products, apps, stays, homes — and every one of them is built around getting past that catalogue's own result ceiling.
4. Risk & Compliance
https://mcp.apify.com/?actors=snow_leo_data/sanctions-screening-api,snow_leo_data/cve-scraper,snow_leo_data/whois-dns-lookup,snow_leo_data/government-tenders-scraper
Four checks an agent runs before it recommends doing business with someone: is the name sanctioned, is the software exploitable, is the domain real and able to receive mail, and is there a public contract behind it.
5. News & Social
https://mcp.apify.com/?actors=snow_leo_data/google-news-scraper,snow_leo_data/telegram-channel-scraper
Two public feeds an agent can read without an account, a bot token or an API key.
How to connect
Claude Desktop and claude.ai
These are remote servers, so they install as custom connectors, not as a config file entry.
- Open Settings (Desktop:
Ctrl+,/ menu → File → Settings. Browser:⌘⇧,). - Click Connectors in the sidebar, then Add → Add custom connector.
- Paste one of the five URLs above and click Add.
- Claude will send you to Apify to sign in. Approve it.
- The tools appear under the connector. You can switch individual tools off in the connector's settings.
Cursor
Put this in ~/.cursor/mcp.json (global) or .cursor/mcp.json (one project).
Add only the servers you want:
{
"mcpServers": {
"snowleo-leads": {
"url": "https://mcp.apify.com/?actors=snow_leo_data/google-maps-scraper,snow_leo_data/yellow-pages-scraper,snow_leo_data/duckduckgo-scraper,snow_leo_data/openstreetmap-business-scraper,snow_leo_data/website-contact-scraper"
},
"snowleo-jobs": {
"url": "https://mcp.apify.com/?actors=snow_leo_data/ats-jobs-scraper,snow_leo_data/seek-scraper,snow_leo_data/jobs-ch-scraper,snow_leo_data/the-muse-jobs-scraper"
},
"snowleo-marketplaces": {
"url": "https://mcp.apify.com/?actors=snow_leo_data/amazon-product-scraper,snow_leo_data/shopify-inventory-scraper,snow_leo_data/app-store-scraper,snow_leo_data/google-play-scraper,snow_leo_data/airbnb-scraper,snow_leo_data/redfin-scraper"
},
"snowleo-risk": {
"url": "https://mcp.apify.com/?actors=snow_leo_data/sanctions-screening-api,snow_leo_data/cve-scraper,snow_leo_data/whois-dns-lookup,snow_leo_data/government-tenders-scraper"
},
"snowleo-media": {
"url": "https://mcp.apify.com/?actors=snow_leo_data/google-news-scraper,snow_leo_data/telegram-channel-scraper"
}
}
}
Cursor will run the OAuth flow the first time the server is used. If you prefer a token to a browser sign-in, add a header instead:
{
"mcpServers": {
"snowleo-leads": {
"url": "https://mcp.apify.com/?actors=snow_leo_data/google-maps-scraper",
"headers": { "Authorization": "Bearer YOUR_APIFY_API_TOKEN" }
}
}
}
Any client that only speaks stdio
Bridge the HTTP endpoint with mcp-remote:
{
"mcpServers": {
"snowleo-leads": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://mcp.apify.com/?actors=snow_leo_data/google-maps-scraper"
]
}
}
}
Checking it by hand
curl -sS -X POST \
"https://mcp.apify.com/?actors=snow_leo_data/telegram-channel-scraper" \
-H "Authorization: Bearer $APIFY_TOKEN" \
-H "Content-Type: application/json" \
-H "Accept: application/json, text/event-stream" \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'
Without a token the same request answers HTTP 401 with
{"error":"invalid_token", ...}. That is expected — the server is gated, not broken.
The gateway publishes OAuth discovery metadata at
https://mcp.apify.com/.well-known/oauth-protected-resource, pointing at Apify
Console as the authorization server, and that server offers dynamic client
registration. Clients that support OAuth therefore need no manual app setup — the
sign-in dialog is all there is.
Tool names
Apify's gateway exposes each Actor as a tool named snow_leo_data--<actor-slug>.
Names longer than 64 characters are truncated and given a short hash suffix, so a
few of the longer Actors appear under a shortened name. The rule lives in
actor_tool_naming.ts
in Apify's open-source gateway.
Alongside our tools, the gateway adds its own helpers — get-actor-run,
get-dataset-items, get-key-value-store-record, abort-actor-run — for
reading back a run that is already finished.
Who pays, and how much
You do, to Apify, with your own account. There is no key to buy from us and nothing to subscribe to. When you connect, you sign in to Apify; your Apify account is billed for the runs your agent starts. Apify's free tier includes monthly platform credit, so small runs often cost nothing out of pocket.
All 21 Actors are cleared by Apify for agentic payments, so an agent can pay for a run on its own without a human in the loop (checked against Apify's public store API on 17 September 2026: 21 of 21).
Every Actor here is priced per result, not per hour or per page. Rows your filters removed, duplicates, and rows already delivered unchanged in monitor mode are not charged. Prices below are the free-plan rate read from Apify's public store API on 17 September 2026; paid Apify plans pay less.
What these tools do not do
Read this before you point an agent at them.
- They read public pages only. Nothing here logs in, buys an account, or passes a paywall. If a source requires a session to show data, that data is not in the output.
- They do not write anything. Every tool is read-only. No posting, no applying to a job, no placing an order, no bidding on a tender.
- They are not instant. A tool call starts an Actor run on Apify. Small runs
finish in seconds; a wide job or map sweep takes one to three minutes. An agent
with a short tool timeout should ask for fewer results, or read the run back
later with
get-dataset-items. - They do not beat the source's own ceiling by magic. Where a ceiling was broken, the number is in the table above and the method is described in the Actor's own page on Apify. Where it was not — Redfin's export cap, Amazon's 306-product window, BBB's 175 — the ceiling is stated, not hidden. A directory is never emptied; it is sampled much more deeply than a plain search allows.
- They do not enrich or guess. A missing website in the OpenStreetMap tool is a missing tag in the data, not a model's inference. Salary in the ATS tool comes from the source field or is absent.
- They are not a database snapshot. Every run reads the source at the moment you run it. Ask the same question twice on a reshuffling source such as Amazon and the two answers will differ.
- No legal advice. The sanctions tool returns what three governments publish. Deciding what that means for a transaction is your compliance team's job, not a language model's.
An honest note about this listing
We publish these servers to MCP directories because agents increasingly find tools there rather than in a marketplace's own search.
We have no measurement that MCP directories bring buyers. Not one. This is a bet on a distribution channel, not a proven channel, and it is described that way on purpose. If it turns out to be empty, this repository will say so.
The condition we set for calling it dead: if, two weeks after the last directory listing goes live, not one of the 21 Actors has gained a single unique user we cannot trace to ourselves, the channel produced nothing and we stop spending time on it.
What is in this repository
README.md this file
LICENSE MIT
set-github-user.sh fills the GitHub namespace placeholder into every server.json
INSTRUCTIONS-RU.md internal publishing notes for the maintainer (Russian)
leads/server.json ─┐
jobs/server.json │ one MCP registry manifest per server,
marketplaces/server.json │ schema 2025-12-11, remote / streamable-http
risk/server.json │
media/server.json ─┘
Each server.json follows the
official registry schema
and all five validate against registry.modelcontextprotocol.io/v0/validate.
They ship with YOUR-GITHUB-USERNAME in the name, repository.url and
websiteUrl fields; ./set-github-user.sh <login> replaces it before publishing.
License
MIT. See LICENSE.